Allow only one indexer node to rollback to 0 at the same time

For HA, index replicas are available. But in cases when rollback to 0 is inevitable, both replicas generally rollback to 0 at the same time. This leads to application unavailability and production outage.

It may be better to allow only 1 indexer node to rollback to 0 at a time. This would make sure that 1 replica copy remains online and can serve the application query requests(though the data may be stale). This can only help if query is not using request_plus consistency and stale results are acceptable.