MDEV-10095 Fix derived tables to return a syntax error instead of "Illegal usage of UNION and LIMIT"

from select_derived_init.

(to simplify further MDEV-8909 changes)

1. Better semantic readability:
- Moving get_select_lex_derived inside select_derived_init
  and decomposing it into get_select_lex and
  $1->init_nested_join(lex->thd)
-  Moving DBUG_ASSERT($1 == Lex->current_select) inside
   select_derived_init

Now init_nested_join() and end_nested_join() reside inside
the same rule select_derived_init.
(It's very likely that they can be further removed,
 as there are no any joins in this rule).

3. Better grammar readability:
  Moving SELECT_SYM from select_derived_init to derived_query_specification.
  It's easier to read a rule when it starts from a terminal symbol.

1. Moving the "| get_select_lex_derived select_derived_init" part of
select_derived into a separate new rule derived_query_specification.
2. Using derived_query_specification directly in select_derived_union
rather than in select_derived.
3. Moving the sequence "opt_order_clause opt_limit_clause opt_select_lock_type"
from select_derived2 to select_derived_union,
after derived_query_specification.

Effectively, the parser now does not go through the sequence
"opt_order_clause opt_limit_clause ... opt_union_order_or_limit" any more.

This fixes the problem with double LIMIT clause and removes 2 shift/reduce
conflicts.

"MDEV-8909 union parser cleanup" changes.

When the server parses a query like
  SELECT * FROM (SELECT 1);
a sequence of calls
  st_select_lex::init_nested_join() ..
  st_select_lex::end_nested_join()
was performed two times (tested in gdb).
Both pairs of calls seem to be redundant for a query
like this, because there are actually no any joins here.

This patch moved "table_ref_select" inside "select_derived",
which revealed that one the pairs was definitely redundant:
After this transformation we got an init_nested_join()
immediately followed by end_nested_join() for the same st_select_lex,
which has no any sense. So this pair of calls was removed.

to simplify further MDEV-8909 refactoring.

This was a bug in the parser. As a result it could accept queries
with invalid derived tables if they used With clauses.

- Moving "SELECT_SYM select_init2_derived opt_table_expression"
  from query_term into a new separate rule query_specification,
  and using query_specification in the beginning of query_term.
- query_term now does not have a %type, query_specification has a %type
  instead. This removes duplicate code that returns
  Lex->current_select->master_unit()->first_select();

between select_derived_union and query_expression_body.
An upcoming patch for MDEV-10035 will also reuse union_head_non_top.

- Moving select_options_and_item_list from select_init2
  to select_init and view_select_aux
- Renaming select_init2 to select_init3
This will simplify upcoming sql_yacc.yy fixes (e.g. MDEV-10035, MDEV-8909).

An addition to original patch:
- use FN_REFLEN instead of HOST_NAME_MAX (the latter can be undefined)
- avoid calling gethostname() on every prompt construction by caching it

Introduce `\H` option which behaves mostly like `\h`. The only exception is
when the client connects to the server hosted on localhost. In this case, the
hostname will be used instead.

closer the grammar in the SQL Standard:
- <query specification> is only a SELECT followed by
  <set quantifier>, <select list> and <table expression>.
- While <query term> includes SELECT queries and derived tables.

MDEV-10036 sql_yacc.yy: Split select_part2 to disallow syntactically bad constructs with INTO, PROCEDURE, UNION
MDEV-10037 UNION with LIMIT ROWS EXAMINED does not require parentheses

MDEV-10030 sql_yacc.yy: Split table_expression and remove PROCEDURE from create_select, select_paren_derived, select_derived2, query_specification

This change refactors the "table_expression" rule in sql_yacc.yy.

Queries with subselects and derived tables, as well as "CREATE TABLE ... SELECT"
now return syntax error instead of "Incorrect usage of PROCEDURE and ...".

This is an addition to original patch. Removed meaningless
"echo SET SESSION AUTOCOMMIT=0" (transaction is started explicitely on
master anyway).

- Don't log connection creation in galera_connect.inc

* simplify the code at default: label
* bugfix: flush the checksum for NULL fields
  (perfschema.checksum was failing)
* cleanup: put repeated code into a function

Error noted by Sergey Vojtovich.

Checksum implementations contain optimizations for calculating
checksums of larger blocks of memory.

This optimization calls my_checksum on a larger block of memory
rather than calling on multiple adjacent memory as its going though
the table columns for each table row.

- Change some static variables to dynamic to ensure that we don't do any memory
  allocations before server starts or stops
- Print more memory information on SIGHUP. Fixed output.
- Write out if memory was lost if run with --debug-at-exit
- Fixed wrong #ifdef in sql_cache.cc