Docs: PCR hub-and-spoke failback known limitation

We have a known limitation in PCR that we would like to publicly document.

In a hub-and-spoke setup, where multiple destinations are streaming from the same source, we do not currently support failing back to the source from one the the destinations once a cutover has occurred. There are edge cases in this scenario that can lead to data consistency issues, so for the time being, we would like to make it explicit that this setup is not supported.

Jira issue: DOC-17225