stop all slices once any slice hits timeout in concurrent search

In concurrent search with timeout in IndexSearcher, when a timeout happens, one slice hits it but still continues processing remaining leaves and keeps failing repeatedly.

Other slices also continue running until they hit the timeout on their own.

This leads to some unnecessary extra work after the timeout.

Would it make sense to stop all slices immediately once any one slice hits the timeout?

Start at IndexSearcher’s concurrent search timeout path and trace how one timed-out slice and the other slices continue processing. Confirm the intended cancellation behavior, then verify that all slices stop after the first timeout without repeated failures or unnecessary extra work.

Written by the indexing model from the issue text.