We must split the creation of table or index metadata from the
creation of the dictionary table records and the creation of
the data. In this way, we can transform a table-rebuilding operation
into an instant ADD COLUMN operation. Dictionary objects will only
be added to cache when table rebuilding or index creation is needed.
The ctx->instant_table will never be added to cache.

innodb_update_n_cols(): Renamed from innobase_update_n_virtual().
Call my_error() if an error occurs.

innobase_add_instant_try(): Call my_error() if an error occurs.

create_index_dict(): Replaces row_merge_create_index_graph().

prepare_inplace_alter_table_dict(): First create the ctx->new_table,
and only then decide if the table really needs to be rebuilt.

dict_create_sys_tables_tuple(), dict_create_table_step(): Omit the
system columns (which will now exist in the dict_table_t object
already at this point).

dict_create_table_step(): Expect the callers to invoke
dict_table_add_system_columns().

pars_create_table(): Before creating the table creation execution graph,
invoke dict_table_add_system_columns().

row_create_table_for_mysql(): Expect all callers to invoke
dict_table_add_system_columns().

dict_table_t::add_to_cache(): Modified from dict_table_add_to_cache().
Let the callers invoke dict_table_add_system_columns() and if needed,
set can_be_evicted.

The field dict_table_t::big_rows was only used for determining if
the adaptive hash index should be used when the internal InnoDB SQL
parser is used. That parser is only used for modifying the InnoDB
data dictionary, updating persistent tables, and for fulltext indexes.

page_cur_search_with_match_bytes(): Skip the comparison of the
MIN_REC_FLAG record. Also, remove some dead code
(we asserted that the search tuple carries no MIN_REC_FLAG).

btr_cur_pessimistic_delete(): Remove the record lock also when the
page is dropped. FIXME: File a bug for this and fix in 5.5+. The bug
is present in all InnoDB versions.

btr_cur_optimistic_update(), btr_cur_pessimistic_update(): Do not trim
the 'default row' record.

innobase_instant_add_try(): Compute and apply the update vector for
the 'default row'.

FIXME: The rollback of the 'default row' operations (that is, the
ROLLBACK of the ALTER TABLE…ADD COLUMN operation) has not been tested
and is unlikely to work. We need a test that flushes the redo log
and kills the server at the end of innobase_instant_add_try(), before
the ALTER TABLE transaction is committed.

btr_cur_instant_init_low(): Read and validate the 'default row'

btr_cur_instant_root_init(): Avoid tripping a debug assertion

innobase_add_instant_try(): Use field_ref_zero for empty values.
Tolerate the hidden FTS_DOC_ID column.

The space management and recovery of operations on off-page columns
relies on the undo log records. Inserting off-page columns during
rollback could break this.

The solution to the rollback problem is simple: Do not duplicate
'default row' data in individual records. In this way, the ROLLBACK
of an UPDATE to a record that was originally written before instant
ADD COLUMN, would never have to insert off-page records, even if
there were instantly added BLOB columns with long DEFAULT values.
This will also reduce space consumption (a simple form of compression).

dtuple_t::trim(index): Omit a redundant suffix of an index tuple right
before insert or update. After instant ADD COLUMN, if the last fields
of a clustered index tuple match the 'default row', there will be no
need to store them. While trimming the entry, we must hold a page latch,
so that the table cannot be emptied and the 'default row' be deleted.

btr_cur_optimistic_update(), btr_cur_pessimistic_update(),
row_upd_clust_rec_by_insert(), row_ins_clust_index_entry_low():
Invoke dtuple_t::trim() if needed.

row_ins_clust_index_entry(): Restore dtuple_t::n_fields after calling
row_ins_clust_index_entry_low().

rec_get_converted_size(), rec_get_converted_size_comp(): Allow the number
of fields to be between n_core_fields and n_fields.

rec_convert_dtuple_to_rec_comp(): Assign the status bits based on the
number of fields.

innobase_add_instant_try(): If the table is empty, remove the 'default row'
instead of adding one.

dict_index_t::commit_instant_copy(): Allow !instant.is_instant() for
the clustered index.

ha_innobase_inplace_ctx::commit_instant(): Allow !instant_table->is_instant()

row_sel_try_search_shortcut_for_mysql(): Skip the 'default row', like row_sel_try_search_shortcut() does

row_upd_changes_field_size_or_external(): Revert some unnecessary
changes, and assert that the update vector does not contain the
magic value UNIV_SQL_DEFAULT.

row_upd_rec_in_place(): Remove a redundant assertion.
rec_set_nth_field() already asserts the same.

row_upd_build_sec_rec_difference_binary(): Assert that there are
no default values in a secondary index tuple.

The ALGORITHM=INPLACE interface is not accessible to temporary tables.
Therefore, instant ADD COLUMN cannot be used either.

dict_table_t::is_temporary(): New predicate.

dict_table_is_temporary(): Define as a wrapper macro.

In the temporary file that is applied by row_log_table_apply(),
we must identify whether the records contain the extra header for
instantly added columns. For now, we will allocate an additional byte
for this for ROW_T_INSERT and ROW_T_UPDATE records when the source table
has been subject to instant ADD COLUMN. The ROW_T_DELETE records are
fine, as they will be converted and will only contain 'core' columns
(PRIMARY KEY and some system columns) that are converted from dtuple_t.

rec_get_converted_size_temp(), rec_init_offsets_temp(),
rec_convert_dtuple_to_temp(): Add the parameter 'status'.

rec_set_n_add_field(): Simplify the interface to resemble
rec_get_n_add_field().

rec_init_offsets_comp_ordinary(): Handle the new rec_leaf_format
constant REC_LEAF_TEMP_COLUMNS_ADDED.

rec_get_converted_size_comp_prefix_low(),
rec_convert_dtuple_to_rec_comp(): Handle
index->is_instant() for both temp and !temp.