Add retries to unregister

[ghostdogpr]

If a pod tries to unregister while the Shard Manager is unavailable, it just logs Shard Manager not available. Can't unregister cleanly and returns. It means the pod will stop without properly unregistering, which can lead to message timeouts.

We should at least retry a few times in case the Shard Manager is just restarting. It can be configurable.