GTID positioning: Ignore filtered domain ids

For setting up a replica from a Galera cluster I would like to be able to set up replication with only the "main" (wsrep_gtid_domain_id) domain configured. The other domain ids (the local gtid_domain_ids that are configured on the cluster nodes) may be ignored, and used at their current position. As these domains are/may be ignored, their history does not matter anyways. I have added them to IGNORE_DOMAIN_ID in the CHANGE MASTER statement.

See related issue MDEV-34485, when replicating from a Galera cluster, the wsrep_gtid_domain_id domain id is the only domain that matters.

When this is implemented, we can easily fail over a replication stream from 1 cluster node to the other, with a load balancer in front of the Galera node.

In this example, I use 'G' for the wsrep_gtid_domain_id, and 'A' for the gtid_domain_id of primary A, 'B' for that of primary B.
The issues that this feature fixes are:

• when failing over to primary B, the only 2 domains known by the primary B are B and G; but the first GTID position of it's local domain A is no longer present in it's binary logs (and vice versa)
• When failing back to node A, the local GTID still known may be too old and no longer present
• If each cluster member uses the same local domain, one node may have a higher position in that domain, and that means duplicate local (usually DDL) statements will cause replication to break (for example MDEV-34492 can cause more local transactions to be logged to one of the cluster nodes

Thank you in advance for taking this into consideration, hopefully some day this logical next step in GTID replication (MariaDB GTID in a Galera cluster) can be added so fail over can occur without human intervention or otherwise hacks.