Using different GTID_DOMAIN_ID than 0 still causes stuck replication if pending queries > slave_parallel_threads

If we execute multiple transactions on primary with a GTID_DOMAIN_ID other than 0 still causes a replication block if more queries completed on primary than there are slave_parallel_threads.

Let's say we have 3 parallel threads.

Let's say we execute
1. set session GTID_DOMAIN_ID=4; alter table x row_format=dynamic;
2. set session GTID_DOMAIN_ID=4; alter table y row_format=dynamic;
3. set session GTID_DOMAIN_ID=4; alter table z row_format=dynamic;

If all three of them complete on primary, while replica is still executing statement 1, other two threads will still stay blocked with message `Waiting for prior transaction to start commit`.

This causes GTID_DOMAIN_ID=0 to not be scheduled on workers and lag starts increasing till all the transactions finish.