[FEATURE REQ]: Parallel write region discovery during per-partition automatic failover.

# Background

During per-partition automatic failover, the backend service takes up to a minute to elect the new write region for the server partition. The Cosmos DB SDK, meanwhile, routes requests in a round-robin manner. This slows down the discovery of the write region for the affected server partition. The goal is to introduce some parallelization in the discovery of the new write region using availability strategy.