[multistage] tracker for query reliability fixes

This is a parent issue to track various issues around multistage query reliability.

- https://github.com/apache/pinot/issues/15997 - graceful restart handling, connection failure detector seems to handle most of this?
- https://github.com/apache/pinot/issues/16253 - adaptive server selection support to route around bad servers
- https://github.com/apache/pinot/issues/17465 - query failures after unhealthy server comes back up (broker <> server)
- https://github.com/apache/pinot/issues/17870 - query failures after unhealthy server comes back up (server <> server)
- https://github.com/apache/pinot/issues/18110 - query failures after broker start up
- https://github.com/apache/pinot/issues/18201 - reduce server fanout w/ pruning
- https://github.com/apache/pinot/pull/18268 - support tcp keepalive to more quickly detect servers that are no longer responding via failure detector
- https://github.com/apache/pinot/pull/18354 - server-side configs for tcp keepalive

Start by reviewing the linked issues covering graceful restarts, adaptive server selection, unhealthy-server recovery, broker startup, and server fanout, along with the linked TCP keepalive pull requests. This is a parent tracker rather than a single implementation task; it is complete when the listed query reliability work has been addressed or otherwise resolved.

Written by the indexing model from the issue text.