Allocator changes to support tiering with replicas

[neetikasinghal]

For a lot of decider checks in AllocationService, while working on replica shards, the corresponding primary shard is required. This can lead to edge cases in tiering, where primary has relocated, but replica has not. The split pool information will only hold one of the shard causing NPE.

To avoid this, one of the solutions is - as part of tiering, we could add a shadowNode to each routing pool. This shadowRoutingNode object, would be a dummy routingNode added only to answer decider queries. It should neither accept shards, nor move shards out.

[peternied]

[Triage - attendees 1 2]
@neetikasinghal Thanks for creating this issue

[linuxpi]

[Storage Triage - attendees 1 2 3 4 5 6 7 8]

@neetikasinghal Thanks for filing this issue, please feel free to submit a pull request! Please add more details to the description and triage priority, sizing and other parameters