Add validation of the total_shards_per_node to the beginning of the ILM shrink step

[joegallo]

If the total_shards_per_node per node setting will prevent the shards of the index from all being relocated to a single node, then we should fail the shrink action without having set-single-node-allocation since it won't work anyway.

For example, if the index has 5 shards, but total_shards_per_node is set to 3 for this index, then we won't be able to actually move all the shards onto one node -- failing this early with a validation / pre-flight error would be better than attempting to set-single-node-allocation and then waiting forever for it to do something that it isn't going to do.

[elasticmachine]

Pinging @elastic/es-core-features (Team:Core/Features)

[joegallo]

Related to #56062 because this could be one of the potential failure cases mentioned there.

[DaveCTurner]

Also relates #44070.

[gmarouli]

Issue #76732 allows the shrink action to unset the index's index.routing.allocation.total_shards_per_node
setting, meaning that there will be no limit and the shrink process can continue. This validation is not necessary anymore.

cc @masseyke , @dakrone