How to support query cache in knn search

[weizijun]

Description

Currently, knn search does not use LRUQueryCache for acceleration. The knn search is heavy when k is big.
Can query caching be supported in knn search?

In semantic search scenario, there are many hot queries. Query caching can speed up these queries.

[sgup432]

@weizijun LRUQueryCache only caches non scoring queries (link), so you can probably check on that first.

In addition, you can look at UsageTrackingQueryCachingPolicy which is the default caching policy, and see if tweaking it for knn queries helps in caching those?