Tap Postgres FastSync inconsistent snapshot for logical replication

[judahrand]

Describe the bug

I believe there is an edge case in the initial snapshot process for Postgres LOG_BASED replication.

This can only occur when snapshotting from a 'replica' and when the replication slot did not previously exist.

Under the above conditions it is possible for the following to occur:

1. a replication slot is created on the primary database
2. the most recent LSN of present on the replica is used as the starting point for logical replication
3. this LSN is actually before the replication slot was created / became consistent due to replication lag between the primary and replica
4. the replica is snapshotted but has missing data that will never be available to the replication slot

I believe this can be solved by #929. This approach is inspired heavily by how Debezium does its initial snapshotting with the addition of waiting for the replica to catch up enough with the primary if needed.

To Reproduce
Steps to reproduce the behavior:

1. Difficult

Expected behavior
I expect all rows in the source database to be present in the destination database.

Screenshots
If applicable, add screenshots to help explain your problem.

Your environment

• Version of PipelineWise, e.g branch/commit #/release/tag
• Source type and setup
• Target type and setup

Additional context
Add any other context about the problem here.