Add config to fail if search targets too many splits

[rdettai-sk]

Description

When a search needs to process too many splits compared to the cluster capacity, we know upfront that it is pretty much impossible for the query to succeed. Instead of trusting resources for the whole duration of the configured timeout, this PR adds a node config to cancel these requests early (after the planning but before sending the jobs to the leafs).

How was this PR tested?

Unit test