Merge new release of InnoDB 5.7.28 to 10.2

MySQL 5.7.28 includes the following InnoDB changes:

Not applicable to MariaDB Server

Bug #29396364 MYSQLD SEGFAULT UNDER LOAD FROM SEVERAL MEMCACHED PLUGIN CLIENTS
MDEV-12050 removed the code related to the Memcached plugin, which was never available in MariaDB Server.

BUG#29862057 MEM0MEM-T FAILS BECAUSE OF NON-INITIALIZED INFRASTRUCTURE FOR EVENTS
This is a follow-up fix related to the MySQL fix of
BUG#29138644 PAGE CLEANER WILL SLEEP FOR LONG TIME IF CLOCK CHANGES
which was fixed differently in MariaDB, in MDEV-14154.

Bug #30150719 INCORRECT LENGTH IN READ_OK_EX()
This is fixing one compiler warning in InnoDB. An equivalent change was part of MDEV-18659.

BUG#29543447: ALTER TABLE ALLOWS TDE ENCRYPTION ON MYISAM ENGINE
This introduces a capability flag to InnoDB, so that the SQL layer can disallow creating encrypted tables in other storage engines than InnoDB. In MariaDB Server, the table options would be used for this.

Bug #29195848 ASSERTION "!OTHER_LOCK" IN LOCK_REC_ADD_TO_QUEUE
This appears to fix a regression that was caused by an earlier change
Bug#25966845 INSERT ON DUPLICATE KEY GENERATE A DEADLOCK
that we did not apply as such in MDEV-19342, because the original problem was partly fixed in MDEV-17073 (and later fully fixed in MDEV-17614).
The intention seems to be to ensure that a partial rollback will not release implicit record locks, but instead convert them into explicit ones. I fail to recognize a reason for such a change.

Applicable to MariaDB Server

Bug #29008298 MYSQLD CRASHES ITSELF WHEN CREATING INDEX
There is no test case, but it looks like the PageBulk code that was added in MySQL 5.7 in WL#7277 is unnecessarily acquiring an index latch and thus delaying concurrent DML or internal tasks, such as the purge of history. The index latch would only be needed after the bulk copying has been completed, to protect dict_index_t::online_log during row_log_apply() or row_log_table_apply(). MDEV-20852 fixes this in a different way.