MDEV-31655: Parallel replication deadlock victim preference code errorneously removed

Restore code to make InnoDB choose the second transaction as a deadlock
victim if two transactions deadlock that need to commit in-order for
parallel replication. This code was erroneously removed when VATS was
implemented in InnoDB.

Also add a test case for InnoDB choosing the right deadlock victim.
Also fixes this bug, with testcase that reliably reproduces:

MDEV-28776: rpl.rpl_mark_optimize_tbl_ddl fails with timeout on sync_with_master

Note: This should be null-merged to 10.6, as a different fix is needed
there due to InnoDB locking code changes.

Signed-off-by: Kristian Nielsen <knielsen@knielsen-hq.org>

sql/sql_insert.cc

@@ -78,6 +78,7 @@
 #include "sql_audit.h"
 #include "sql_derived.h"                        // mysql_handle_derived
 #include "sql_prepare.h"
+#include "rpl_rli.h"
 #include <my_bit.h>
 
 #include "debug_sync.h"
@@ -1753,6 +1754,12 @@ int write_record(THD *thd, TABLE *table,COPY_INFO *info)
   save_read_set=  table->read_set;
   save_write_set= table->write_set;
 
+  DBUG_EXECUTE_IF("rpl_write_record_small_sleep_gtid_100_200",
+    {
+      if (thd->rgi_slave && (thd->rgi_slave->current_gtid.seq_no == 100 ||
+                             thd->rgi_slave->current_gtid.seq_no == 200))
+        my_sleep(20000);
+    });
   if (info->handle_duplicates == DUP_REPLACE ||
       info->handle_duplicates == DUP_UPDATE)
   {