Also adjusted some test files.

Due to inconsistent usage of different cost models to calculate
the cost of ref accesses we have to make the calculation of the
gain promising by usage a range filter more complex.

The code was rewritten in the same way as the code of
ha_partition::multi_range_read_info_const() had been rewritten
earlier.

The fix allowed to run spider.partition_mrr.

Find indexes of one table which parts participate in one constraint.
These indexes are called constraint correlated.

New methods: TABLE::find_constraint_correlated_indexes() and
virtual method check_index_dependence() were added.
For each index it's own constraint correlated index map was created
where all indexes that are constraint correlated with the current are
marked.

The results of this task are used for MDEV-16188 (Use in-memory
PK filters built from range index scans).

Fix bugs caused by changes for ANALYZE

Also adjusted some result files after Galina's last patch for ANALYZE.

ANALYZE and ANALYZE FORMAT=JSON structures are changed in the way that they
show additional information when rowid filter is used:

- r_selectivity_pct - the observed filter selectivity
- r_buffer_size - the size of the rowid filter container buffer
- r_filling_time_ms - how long it took to fill rowid filter container

New class Rowid_filter_tracker was added. This class is needed to collect data
about how rowid filter is executed.

Also pushed new results for regression_mdev6282.test.
(the cause of difference should be investigated).

They generally don't have stats.records=0, the the statistics are computed
based on GetApproximateSizes and GetApproximateMemTableStats. Still, when
there's nothing on disk and only a few records are in the MemTable, it
is possible that both calls will return 0, and the optimizer will see
stats.records=0, which may cause "impossible range" condition (even if
records_in_range() returned a non-zero value)

Fix the formatting, and remove the MONITOR interface.
Remove unnecessary wrapper functions for the callbacks,
and replace void* with ha_innobase*.

This patch contains a full implementation of the optimization
that allows to use in-memory rowid / primary filters built for range  
conditions over indexes. In many cases usage of such filters reduce  
the number of disk seeks spent for fetching table rows.

In this implementation the choice of what possible filter to be applied  
(if any) is made purely on cost-based considerations.

This implementation re-achitectured the partial implementation of
the feature pushed by Galina Shalygina in the commit
8d5a1112.

Besides this patch contains a better implementation of the generic  
handler function handler::multi_range_read_info_const() that
takes into account gaps between ranges when calculating the cost of
range index scans. It also contains some corrections of the
implementation of the handler function records_in_range() for MyISAM.

This patch supports the feature for InnoDB and MyISAM.

add tests

Add a system variable spider_slave_trx_isolation.
- spider_slave_trx_isolation
  The transaction isolation level when Spider table is used by slave SQL thread.
  -1 : OFF
   0 : READ UNCOMMITTED
   1 : READ COMMITTED
   2 : REPEATABLE READ
   3 : SERIALIZABLE
  The default value is -1

Miscellaneous Spider typos

Change default value of the followings
quick_mode 0 -> 3
quick_page_size 100 -> 1024

Add the following parameter for limiting result page size by byte
- quick_page_byte(qpb)
  Number of bytes in a page when acquisition one by one.
  When quick_mode is 1 or 2, Spider stores at least 1 record even if
  quick_page_byte is smaller than 1 record. When quick_mode is 3,
  quick_page_byte is used for judging using temporary table.
  That is given to priority when server parameter spider_quick_page_byte
  is set.
  The default value is 10485760

Fix "out of sync" issue at using quick_mode = 1 or 2

The fields of the temporary table were not created in create_tmp_table function. Because item->const_item() was true. But the temporary tables that is created by Spider are always used all columns. So Spider should call create_tmp_table function with TMP_TABLE_ALL_COLUMNS flag.

The default value for eq_range_index_dive_limit is set to 10

Replaced wait on COND_thread_count with busy waiting with 1 millisecond
sleep.

Aim is to reduce usage of LOCK_thread_count and COND_thread_count.

Signal handler is now respoinsible for setting abort_loop and breaking
poll() in main thread. The rest is handled by main thread itself.

Removed redundant LOCK_error_log init/destroy wrappers.
Removed redundant unireg_end(): it is trivial and it has only one caller.
Removed unused ready_to_exit from PFS.
Removed kill_in_progress: duplicates abort_loop.
Removed shutdown_in_progress: duplicates abort_loop.
Removed ready_to_exit: was used to make sure main thread waits for
cleanups, which are now done by main thread itself.
Removed SIGNALS_DONT_BREAK_READ, MAYBE_BROKEN_SYSCALL,
kill_broken_server: never defined/used.
Make clean_up() static.

fill_status.

Also, remove LOCK_status around calc_sum_of_all_status()

Also, rename LOCK_show_status into LOCK_all_status_vars.
This reflects the variable the lock protects.

modifications (insert/erase) are protected by write lock
iteration over list is protected by read lock.

This way, threads that iterate over the list (as in SHOW PROCESSLIST,
SHOW GLOBAL STATUS) do not block each other.

* fix CRL tests to work
* regenerate certificates to be at least 2048 bit
  (fixes buster and rhel8 in buildbot)
* update generate-ssl-cert.sh to generate crl files
* make all SSL tests to use certificates generated
  in generate-ssl-cert.sh, remove unused certificates

from info to debug

In contrast to thread_count, which is decremented by THD destructor,
this one was most probably intended to be decremented after all THD
destructors are done.

THD_count class was added to achieve similar effect with thread_count.

Aim is to reduce usage of LOCK_thread_count and COND_thread_count.
Part of MDEV-15135.

Implemented and integrated THD_list as a replacement for the global
thread list. It uses own mutex instead of LOCK_thread_count for THD
list protection.

Removed unused first_global_thread() and next_global_thread().

delayed_insert_threads is now protected by LOCK_delayed_insert. Although
this patch doesn't fix very wrong synchronization of this variable.

After this patch there are only 2 legitimate uses of LOCK_thread_count
left, both in mysqld.cc: thread_count and ready_to_exit.

Aim is to reduce usage of LOCK_thread_count and COND_thread_count.
Part of MDEV-15135.

LOG_INFO::lock was useless. It could've only protect against concurrent
iterators execution, which was already protected by LOCK_thread_count.

Use LOCK_thd_data instead of LOCK_thread_count as a protection against
THD::current_linfo reset.

Aim is to reduce usage of LOCK_thread_count and COND_thread_count.
Part of MDEV-15135.

Bootstrap in a separate thread was introduced in 746f0b3b to workaround
OS/2 small stack size. OS/2 support was discontinued in 2006 and modern
operating systems have default stack size a few times larger than
default thread_stack and it is tunable.

Aim is to reduce usage of LOCK_thread_count and COND_thread_count.
Part of MDEV-15135.