Signalling table not working for SQLServer if snapshot comes from a replica [DBZ-3604]

Migrated from [DBZ-3604](https://issues.redhat.com/browse/DBZ-3604)

On environments where the impact of the snapshot is not affordable to be done in the primary the obvious option is create a secondary and perform the snapshot there without impact productions environments.

The current implementation of the signaling assumes that we are always running the connector against a SQLServer that is writable.

It would be interesting if this signaling table could be place in a different database that allow us to have read-write permissions for signalling and the main database pointed to a readOnly replica for perform the snapshot at the same time.

Start by tracing the SQL Server signaling-table implementation and how the connector configures its database connection during snapshots. Confirm the desired design for writing signals to a writable database while reading snapshot data from a read-only replica, then verify signaling and snapshot behavior in that setup.

Written by the indexing model from the issue text.