Query cache used for results of vector search conflicts with the purpose of mhnsw_min_limit

Currently results of vector search queries are written into the query cache and read from it.
It means that increasing mhnsw_min_limit at runtime for the purpose of a more precise search won't have any effect, as the previous less precise results will be still returned from the cache.