[Segment Replication] Handle new ReplicationCheckpoints when received by replicas.

[mch2]

With issue #2927, ReplicationCheckpoints will be received and passed to IndexShards for processing.

Shards should receive these requests and:

1. Compute the existing checkpoint and validate against the requested checkpoint. Dedupe incoming requests & ensure we are processing one checkpoint at a time.
2. Process the checkpoint by kicking off a new replication event.
3. Handle replication failure.

A version of this logic is implemented in the feature branch here, but needs to be revisited and PR'd to main. This is an effort to cut the merge into smaller digestible chunks. This can trigger an empty start method inside of a service class.