this is a cleanup patch for our current auto_increment handling:
new names for auto_increment variables in THD, new methods to manipulate them
(see sql_class.h), some move into handler::, causing less backup/restore
work when executing substatements. 
This makes the logic hopefully clearer, less work is is needed in
mysql_insert().
By cleaning up, using different variables for different purposes (instead
of one for 3 things...), we fix those bugs, which someone may want to fix
in 5.0 too:
BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate
statement-based"
BUG#20341 "stored function inserting into one auto_increment puts bad
data in slave"
BUG#19243 "wrong LAST_INSERT_ID() after ON DUPLICATE KEY UPDATE"
(now if a row is updated, LAST_INSERT_ID() will return its id)
and re-fixes:
BUG#6880 "LAST_INSERT_ID() value changes during multi-row INSERT"
(already fixed differently by Ramil in 4.1)
Test of documented behaviour of mysql_insert_id() (there was no test).
The behaviour changes introduced are:
- LAST_INSERT_ID() now returns "the first autogenerated auto_increment value
successfully inserted", instead of "the first autogenerated auto_increment
value if any row was successfully inserted", see auto_increment.test.
Same for mysql_insert_id(), see mysql_client_test.c.
- LAST_INSERT_ID() returns the id of the updated row if ON DUPLICATE KEY
UPDATE, see auto_increment.test. Same for mysql_insert_id(), see
mysql_client_test.c.
- LAST_INSERT_ID() does not change if no autogenerated value was successfully 
inserted (it used to then be 0), see auto_increment.test.
- if in INSERT SELECT no autogenerated value was successfully inserted,
mysql_insert_id() now returns the id of the last inserted row (it already
did this for INSERT VALUES), see mysql_client_test.c.
- if INSERT SELECT uses LAST_INSERT_ID(X), mysql_insert_id() now returns X
(it already did this for INSERT VALUES), see mysql_client_test.c.
- NDB now behaves like other engines wrt SET INSERT_ID: with INSERT IGNORE,
the id passed in SET INSERT_ID is re-used until a row succeeds; SET INSERT_ID
influences not only the first row now.

Additionally, when unlocking a table we check that the thread is not keeping
a next_insert_id (as the table is unlocked that id is potentially out-of-date);
forgetting about this next_insert_id is done in a new
handler::ha_release_auto_increment().

Finally we prepare for engines capable of reserving finite-length intervals
of auto_increment values: we store such intervals in THD. The next step
(to be done by the replication team in 5.1) is to read those intervals from
THD and actually store them in the statement-based binary log. NDB
will be a good engine to test that.


mysql-test/extra/binlog_tests/binlog.test:
  Testing that if INSERT_ID is set to a value too big for the
  column's type, the binlogged INSERT_ID is the truncated value
  (important if slave has a column of a "wider" numeric type).
  Testing binlogging of INSERT_ID with INSERT DELAYED, to be sure that 
  we binlog an INSERT_ID event only for the delayed rows which use one.
mysql-test/extra/rpl_tests/rpl_insert_id.test:
  Testcase for BUG#20339 "stored procedure using
  LAST_INSERT_ID() does not replicate statement-based".
  Testcase for BUG#20341 "stored function inserting into one
  auto_increment puts bad data in slave".
mysql-test/extra/rpl_tests/rpl_loaddata.test:
  Test that LOAD DATA INFILE sets a value for a future LAST_INSERT_ID().
mysql-test/r/auto_increment.result:
  behaviour change: when INSERT totally fails (not even succeeds
  partially and then rolls back), don't change last_insert_id().
  Behaviour change: LAST_INSERT_ID() is now the first successfully inserted,
  autogenerated, id.
  Behaviour change: if INSERT ON DUPLICATE KEY UPDATE, if the table has auto_increment
  and a row is updated, then LAST_INSERT_ID() returns the id of this row.
mysql-test/r/binlog_row_binlog.result:
  result update
mysql-test/r/binlog_stm_binlog.result:
  result update
mysql-test/r/insert.result:
  result update
mysql-test/r/rpl_insert_id.result:
  result update
mysql-test/r/rpl_loaddata.result:
  result update
mysql-test/r/rpl_ndb_auto_inc.result:
  ndb's behaviour is now like other engines wrt SET INSERT_ID
  in a multi-row INSERT:
  - with INSERT IGNORE: the id passed in SET INSERT_ID is re-used until
  a row succeeds.
  - generally, SET INSERT_ID sets the first value and other values are
  simply computed from this first value, instead of previously where
  the 2nd and subsequent values where not influenced by SET INSERT_ID;
  this good change is due to the removal of "thd->next_insert_id=0"
  from ha_ndbcluster.
mysql-test/t/auto_increment.test:
  A testcase of BUG#19243: if ON DUPLICATE KEY UPDATE updates a row,
  LAST_INSERT_ID() now returns the id of the row.
  Test of new behaviour of last_insert_id() when no autogenerated value was
  inserted, or when only some autogenerated value (not the first of them) was
  inserted.
mysql-test/t/insert.test:
  testing INSERT IGNORE re-using generated values
sql/ha_federated.cc:
  update for new variables.
sql/ha_ndbcluster.cc:
  handler::auto_increment_column_changed not needed, equivalent to
  (insert_id_for_cur_row > 0).
  thd->next_insert_id=0 not needed anymore; it was used to force
  handler::update_auto_increment() to call ha_ndbcluster::get_auto_increment()
  for each row of a multi-row INSERT, now this happens naturally
  because NDB says "I have reserved you *one* value" in get_auto_increment(),
  so handler::update_auto_increment() calls again for next row.
sql/handler.cc:
  More comments, use of new methods and variables. Hopes to be clearer
  than current code.
  thd->prev_insert_id not in THD anymore: it is managed locally by inserters
  (like mysql_insert()).
  THD::clear_next_insert_id is now equivalent to
  handler::next_insert_id > 0.
  get_auto_increment() reserves an interval of values from the engine,
  uses this interval for next rows of the statement, until interval
  is exhausted then it asks for another interval (of a bigger size
  than the first one; size doubles until reaching 65535 then it stays constant).
  If doing statement-based binlogging, intervals are remembered in a list
  for storage in the binlog.
  For "forced" insert_id values (SET INSERT_ID or replication slave),
  forced_auto_inc_intervals is non-empty and the handler takes its intervals
  from there, without calling get_auto_increment().
  ha_release_auto_increment() resets the handler's auto_increment variables;
  it calls release_auto_increment() which is handler-dependent and
  serves to return to the engine any unused tail of the last used
  interval.
  If ending a statement, next_insert_id>0 means that autoinc values have been
  generated or taken from the master's binlog (in a replication slave) so
  we clear those values read from binlog, so that next top- or sub-
  statement does not use them.
sql/handler.h:
  handler::auto_increment_changed can be replaced by
  (handler::insert_id_for_cur_row > 0).
  THD::next_insert_id moves into handler (more natural, and prepares
  for the day when we'll support a single statement inserting into
  two tables - "multi-table INSERT" like we have UPDATE - will this
  happen?).
  This move makes the backup/restore of THD::next_insert_id when entering
  a substatement unneeded, as each substatement has its own handler
  objects.
sql/item_func.cc:
  new names for variables.
  For the setting of what mysql_insert_id() will return to the client,
  LAST_INSERT_ID(X) used to simply pretend that the generated autoinc
  value for the current row was X, but this led to having no reliable
  way to know the really generated value, so we now have a bool:
  thd->arg_of_last_insert_id_function which enables us to know that
  LAST_INSERT_ID(X) was called (and then X can be found in
  thd->first_successful_insert_id_in_prev_stmt).
sql/log.cc:
  new variable names for insert_ids. Removing some unused variables in the slow
  log.
sql/log_event.cc:
  new variable names, comments. Preparing for when master's won't binlog
  LAST_INSERT_ID if it was 0.
sql/set_var.cc:
  new variable names.
  The last change repeats how Bar fixed BUG#20392
  "INSERT_ID session variable has weird value" in 5.0.
sql/sql_class.cc:
  new variables for insert_id. In THD::cleanup_after_query() we fix
  BUG#20339 "stored procedure using LAST_INSERT_ID() does not replicate
  statement-based" (will one want to fix it in 5.0?). Many comments
  about what stored functions do to auto_increment.
  In reset|restore_sub_statement_state(), we need to backup less
  auto_inc variables as some of them have moved to the handler;
  we backup/restore those which are about the current top- or sub-
  statement, *not* those about the statement-based binlog
  (which evolve as the top- and sub-statement execute).
  Because we split THD::last_insert_id into 
  THD::first_successful_insert_id_in_prev_stmt and
  THD::auto_inc_intervals_for_binlog (among others), we fix
  BUG#20341 "stored function inserting into one auto_increment
  puts bad data in slave": indeed we can afford to not backup/restore
  THD::auto_inc_intervals_for_binlog (which fixes the bug) while still
  backing up / restoring THD::first_successful_insert_id_in_prev_stmt
  (ensuring that the top-level LAST_INSERT_ID() is not affected by INSERTs
  done by sub-statements, as is desirable and tested in rpl_insert_id.test).
sql/sql_class.h:
  new variables and methods for auto_increment.
  Some THD members move into handler (those which are really about
  the table being inserted), some stay in THD (those which are
  about what a future LAST_INSERT_ID() should return, or about
  what should be stored into the statement-based binlog).
  THD::next_insert_id moves to handler::.
  THD::clear_next_insert_id removed (had become equivalent
  to next_insert_id > 0).
  THD::last_insert_id becomes four:
  THD::first_successful_insert_id_in_cur_stmt,
  THD::auto_inc_intervals_for_binlog,
  handler::insert_id_for_cur_row,
  THD::first_successful_insert_id_in_prev_stmt.
  THD::current_insert_id becomes:
  THD::first_successful_insert_id_in_prev_stmt_for_binlog
  THD::prev_insert_id is removed, handler can just use
  handler::insert_id_for_cur_row instead (which is more accurate:
  for the first row, prev_insert_id was set before get_auto_increment
  was called, so was 0, causing a call to
  get_auto_increment() for the 2nd row if the 1st row fails;
  here we don't need the call as insert_id_for_cur_row has
  the value of the first row).
  THD::last_insert_id_used becomes: stmt_depends_on_first_row_in_prev_stmt
  THD::insert_id_used is removed (equivalent to
  auto_inc_intervals_for_binlog non empty).
  The interval returned by get_auto_increment() and currently being
  consumed is handler::auto_inc_interval_for_cur_row.
  Comments to explain each of them.
  select_insert::last_insert_id becomes autoinc_value_of_last_inserted_row.
sql/sql_insert.cc:
  the "id" variable is not changed for each row now; it used to compensate for
  this contradiction:
  - thd->last_insert_id supposed job was to keep the id of the first row
  - but it was updated for every row
  - so mysql_insert() made sure to catch its first value and restore it at the end of stmt.
  Now THD keeps the first value in first_successful_insert_id_in_cur_stmt,
  and value of the row in insert_id_for_cur_row. So "id" only serves to fill
  mysql_insert_id(), as depending on some conditions, "id" must be different
  values.
  Prev_insert_id moves from THD to write_record().
  We now set LAST_INSERT_ID() in ON DUPLICATE KEY UPDATE too (BUG#19243).
  In an INSERT DELAYED, we still "reset auto-increment caching" but differently
  (by calling ha_release_auto_increment()).
sql/sql_load.cc:
  no need to fiddle with "id", THD maintains
  THD::first_successful_insert_id_in_cur_stmt by itself and correctly now.
  ha_release_auto_increment() is now (logically) called before we unlock
  the table.
sql/sql_parse.cc:
  update to new variable names.
  Assertion that reset_thd_for_next_command() is not called for every
  substatement of a routine (I'm not against it, but if we do this change,
  statement-based binlogging needs some adjustments).
sql/sql_select.cc:
  update for new variable names
sql/sql_table.cc:
  next_insert_id not needed in mysql_alter_table(), THD manages.
sql/sql_update.cc:
  update for new variable names.
  Even though this is UPDATE, an insert id can be generated (by
  LAST_INSERT_ID(X)) and should be recorded because mysql_insert_id() wants
  to know about it.
sql/structs.h:
  A class for "discrete" intervals (intervals of integer numbers with a certain
  increment between them): Discrete_interval, and a class for a list of such
  intervals: Discrete_intervals_list
tests/mysql_client_test.c:
  tests of behaviour of mysql_insert_id(): there were no such tests, while in
  our manual we document its behaviour. In comments you'll notice the behaviour
  changes introduced (there are 5).