Description
I think we have two issues with LRUQueryCache.CachingWrapperWeight:
- when generating cached entries, it creates a Scorer and loads it into a DocIdSet while it would be more efficient to use a BulkScorer to do so since we consume all matching documents
- when a query should not be cached (eg. because it has not be seen often enough yet), we do not delegate the bulk scorer, which means that we will use a DefaultBulkScorer on top of the scorer, which can be significantly slower for some queries (eg. disjunctions)