PreQueueingHint Extension Point for Scheduler Event Processing

This KEP introduces a PreQueueingHintFn extension point to the Kubernetes scheduling framework. When a cluster event occurs, the scheduler currently iterates all unschedulable pods to evaluate per-pod QueueingHintFn. The new PreQueueingHintFn is called once per event before this iteration, allowing plugins to return a targeted list of pods to check — or signal that no pods need checking. This reduces the requeue path from O(N) to O(1) for events that affect only specific pods. The DRA plugin implements PreQueueingHintFn for ResourceClaim events using a pod informer index that maps ResourceClaims to the pods referencing them.