Let us avoid any overhead in release builds, for an empty function.

The merge commit 0fd89a1a
of commit b6ec1e8b
was slightly incomplete.

ReadView::mem_valid(): Use the correct primitive
MEM_MAKE_ADDRESSABLE(), because MEM_UNDEFINED() now has
no effect on ASAN.

recv_sys_t::alloc(), recv_sys_t::add(): Use MEM_MAKE_ADDRESSABLE()
instead of MEM_UNDEFINED(), to get the correct behaviour for ASAN.
For Valgrind and MSAN, there is no change in behaviour.

recv_sys_t::free(), recv_sys_t::clear(): Before freeing memory to
buf_pool.free_list, invoke MEM_MAKE_ADDRESSABLE() on the entire
buf_block_t::frame, to cancel the effect of MEM_NOACCESS() in
recv_sys_t::alloc().

first try discovering engines, then the rest.
otherwise every DROP TABLE non_existent; will do
lots of i/o trying to remove .MYI/.MYD/.MAI/.MAD/.CSV/etc files

this matches the old behavior where DROP TABLE always tried to discover
the table before dropping.

normal DROP TABLE with many tables continues after an error,
trying to drop as many tables as possible. But DROP TEMPORARY TABLE
was aborting on the first error. Change it to behave as DROP TABLE does.

don't do table discovery on DROP. DROP falls back to "force"
approach when a table isn't found and will try to drop in all
engines anyway. That is, trying to discover in all engines before
the drop is redundant and may be expensive.

drop-no_root needs DROP DATABASE to fail.
But `chmod 000` is not reliable after drop table force anymore.

Make DROP DATABASE to fail by creating an extra file in the db dir

remove redundant argument, return all possible enum values

minor post-review fixes
* remove duplicate tests
* first file indicates table existance even with discovery
* don't duplicate trigger dropping code

first step in moving drop table out of the handler.
todo: other methods that don't need an open table

for now hton->drop_table is optional, for backward compatibility
reasons

include/my_valgrind.h:88:112: error: ‘void* memset(void*, int, size_t)’ writing to an object of non-trivial type ‘key_map’ {aka ‘class Bitmap<64>’}; use assignment instead [-Werror=class-memaccess]

in this case it's safe, Bitmap<> is trivial enough

In AddressSanitizer, we only want memory poisoning to happen
in connection with custom memory allocation or freeing.

The primary use of MEM_UNDEFINED is for declaring memory uninitialized
in Valgrind or MemorySanitizer. We do not want MEM_UNDEFINED to
have the unwanted side effect that AddressSanitizer would no longer
be able to complain about accessing unallocated memory.

MEM_UNDEFINED(): Define as no-op for AddressSanitizer.

MEM_MAKE_ADDRESSABLE(): Define as MEM_UNDEFINED() or
ASAN_UNPOISON_MEMORY_REGION().

MEM_CHECK_ADDRESSABLE(): Wrap also __asan_region_is_poisoned().

- Some of the bug fixes are backports from 10.5!
- The fix in innobase/fil/fil0fil.cc is just a backport to get less
  error messages in mysqld.1.err when running with valgrind.
- Renamed HAVE_valgrind_or_MSAN to HAVE_valgrind

Fixed by:
- Make all quick_* variable allocated according to real number keys instead
  of MAX_KEY
- Store all the quick* items in separated allocated structure (OPT_RANGE)
- Ensure we don't access any quick* variable without first checking
  opt_range_keys.is_set().  Thanks to this, we don't need any
  pre-initialization of quick* variables anymore.

Some renames was done to use the new structure:
table->quick_keys                -> table->opt_range_keys
table->quick_rows[X]             -> table->opt_range[X].rows
table->quick_key_parts[X]        -> table->opt_range[X].key_parts
table->quick_costs[X]            -> table->opt_range[X].cost
table->quick_index_only_costs[X] -> table->opt_range[X].index_only_cost
table->quick_n_ranges[X]         -> table->opt_range[X].ranges
table->quick_condition_rows      -> table->opt_range_condition_rows

This patch should both decrease memory needed for TABLE objects
(3528 -> 984 + keyinfo) and increase performance, thanks to less
initializations per query, and more localized memory, thanks to the
opt_range structure.

When using field_conv(), which is called in case of field1=field2 copy in
fill_records(), full varstring's was copied, including unitialized bytes.
This caused valgrind to compilain about usage of unitialized bytes when
using Aria static length records.
Fixed by not using memcpy when copying varstrings but instead just copy
the real bytes.

- Removed not needed bzero in void TABLE::initialize_quick_structures().
- Replaced bzero with TRASH_ALLOC() to have this change verfied with
  memory checkers
- Added missing table->quick_keys.is_set in table_cond_selectivity()

MDEV-15053 inadvertently inverted the condition for invoking
buf_read_ahead_linear().

buf_page_get_low(): Apply the correct condition.

buf_page_optimistic_get(): Remove the read-ahead altogether.
This function should never access a block for the first time.

This reverts commit e0793d38.

In idiomatic C++, accessor functions should not discard qualifiers.