TVC can be used in IN subquery and in PARTITION BY struct now.
Special variable to control working of optimization added.

Trasformation goes that way:
a in (1,2) ->
a in (select * from (values ((1),(2))) as new_tvc)

Special variable that controlls optimization added.
Now optimization works only in several cases.
New tests added.

Old tests corrected. Now with TVC explain can be used. TVC also can be used in recursive CTEs.

TVCs can be used separately, with UNION/UNION ALL, in derived tables,
in views and in common table expressions.

Tests corrected.
Test results added.

TVC can be used in UNION-statement, in view and in subquery.

Files where TVC is defined and its methods are stored added.
Methods exec and prepare for TVC added.
Tests for TVC added.

CLOCKS_PER_SEC not used anymore.
start_timer()/mysql_end_timer() don't make much sense.
Buffer size requirement increased by 1 byte.

As svoj points out, my_timer_microseconds uses gettimeofday which is
affected by "discontinuous jumps in the system time", according to
man.

This patch changes to use microsecond_interval_timer which is "not
perfect, but is less affected by these jumps", and also does not
require an include of my_rdtsc.h header.

https://github.com/MariaDB/server/pull/332#discussion_r114708923

"No space before "=" and space after cast please"
https://github.com/MariaDB/server/pull/332#discussion_r114708923

Oversight spotted by svoj:
https://github.com/MariaDB/server/pull/332#discussion_r114708923

based upon: https://github.com/dveeden/mysql-server/commit/d5e46428075d86dbdc333d27951bb06cb4c11a32

MariaDB [test]> select sleep(0.123);
+--------------+
| sleep(0.123) |
+--------------+
|            0 |
+--------------+
1 row in set (0.123 sec)

"More exact timing for mysql client based on my_timer_microseconds"

Based on suggestion from @grooverdan on https://github.com/mysql/mysql-server/pull/112

This patch is slightly bigger because the original did not preserve the
return type of my_timer_microseconds and this patch does.
   (my_timer_microseconds returns ulonglong, not simply ulong)

Also I believe the correct place to do the division of microseconds to
seconds is better in the caller of nice_time because nice_time takes a
"double sec" param, so we should convert before calling; the other caller
of nice_time does not call with microseconds.

The following options will be removed:

innodb_file_format
innodb_file_format_check
innodb_file_format_max
innodb_large_prefix

They have been deprecated in MySQL 5.7.7 (and MariaDB 10.2.2) in WL#7703.

The file_format column in two INFORMATION_SCHEMA tables will be removed:

innodb_sys_tablespaces
innodb_sys_tables

Code to update the file format tag at the end of page 0:5
(TRX_SYS_PAGE in the InnoDB system tablespace) will be removed.
When initializing a new database, the bytes will remain 0.

All references to the Barracuda file format will be removed.
Some references to the Antelope file format (meaning
ROW_FORMAT=REDUNDANT or ROW_FORMAT=COMPACT) will remain.

This basically ports WL#7704 from MySQL 8.0.0 to MariaDB 10.3.1:

commit 4a69dc2a95995501ed92d59a1de74414a38540c6
Author: Marko Mäkelä <marko.makela@oracle.com>
Date:   Wed Mar 11 22:19:49 2015 +0200

This excludes MDEV-12472 (InnoDB should accept XtraDB parameters,
warning that they are ignored). In other words, MariaDB 10.3 will not
recognize any XtraDB-specific parameters.

Fixed the following things from the above MDEV:
- Ensure the user has INSERT privilege when generating new sequence values
  with NEXT VALUE FOR or SETVAL()
- Fixed bug in InnoDB when generating several sequence values in one statement
- Ensure that read_set is up to date before calling ha_sequence::ha_write_row()
  - This is only a potential bug with storage engines that trusts the column maps completely

This is needed for MyISAM and other storage engines which normally
relies on THR_LOCK's to ensure that one is not writing the same block
one could be reading from.

The fix for MDEV-12866 fixed this problem as well. Adding tests only.

Recent changes in Type_handler fixed this problem. Adding tests only.

Adding tests for MDEV-9408 CREATE TABLE SELECT MAX(int_column) creates different columns for table vs view

Recent fixes in Type_handler fixed this problem. Adding tests only.

Adding tests for MDEV-9406 CREATE TABLE..SELECT creates different columns for IFNULL() and equivalent COALESCE,CASE,IF

Recent changes in Type_handler fixed this problem as well.
Adding tests only.

Previous fixes in Type_handler
(see the patch for MDEV-12875, MDEV-12886, MDEV-12916)
fixed this problem as well. Adding only the test from the report.

Fixing to pass item->name.str rather than item->name to my_error().

row_search_mvcc(): Relax debug assertions. For NO_ROLLBACK tables,
a transaction or a read view will not be started at the start of
a statement.

row_upd_clust_step(): Support CREATE TEMPORARY SEQUENCE or
CREATE TEMPORARY TABLE…NO_ROLLBACK.

A cleanup for MDEV-11514, MDEV-11497, MDEV-11554, MDEV-11555 - IN and CASE type aggregation problems

Removing cmp_item::get_comparator() and calling instead
Type_handler::make_cmp_item(), which was earlier introduced by this patch:

74891ed2