[Resharding] Retry COPY on transient errors

[levkk]

Description

Retry parallel table copy if an error that's easily recoverable happens:

• can't connect to server (temporarily down, network issue, etc.)
• copy interrupted due to network error (server restarted mid copy); this is safe because copies are transactional.

The latter needs to handle two situations:

1. Destination shard is down: just retry
2. Origin is down: need to re-start the transaction and re-create the temporary replication slot