MDEV-14984 - regression in connect performance

Removed redundant plugin_thdvar_cleanup() from end_connection(): called by
THD::free_connection(), which always follows end_connection().

Saves at least one lock(LOCK_plugin) and one
rdlock(LOCK_system_variables_hash).

Benchmarked on a 2socket/20core/40threads Broadwell system using sysbench
connect brencmark @40 threads (with select 1 disabled).

10.2 shows moderate improvement: 136219.93 -> 137766.31 CPS.
10.3 is improvement is somewhat better: 93018.29 -> 101379.77 CPS.

Also backported MyRocks memory leak fix from 10.4, which turned out to
be unrelated.

include/thread_pool_priv.h

@@ -98,8 +98,6 @@ void mysql_audit_release(THD *thd);
 bool thd_is_connection_alive(THD *thd);
 /* Close connection with possible error code */
 void close_connection(THD *thd, uint errcode);
-/* End the connection before closing it */
-void end_connection(THD *thd);
 /* Decrement connection counter */
 void dec_connection_count();
 /* Destroy THD object */

sql/sql_connect.cc

@@ -1122,7 +1122,6 @@ void end_connection(THD *thd)
   }
   thd->wsrep_client_thread= 0;
 #endif
-  plugin_thdvar_cleanup(thd);
 
   if (thd->user_connect)
   {

storage/rocksdb/ha_rocksdb.cc

@@ -14010,6 +14010,8 @@ rocksdb_set_update_cf_options(THD *const /* unused */,
 
   RDB_MUTEX_LOCK_CHECK(rdb_sysvars_mutex);
 
+  my_free(*reinterpret_cast<void **>(var_ptr));
+
   if (!val) {
     *reinterpret_cast<char **>(var_ptr) = nullptr;
     RDB_MUTEX_UNLOCK_CHECK(rdb_sysvars_mutex);