Is your feature request related to a problem? Please describe.
This issue relates to using Milvus efficiently. It becomes important as the size of the cache grows.

Describe the solution you'd like
This issue concerns the FindSimilar function for the Milvus cache backend in milvus_cache.go. Milvus is optimized for efficient vector comparisons. Therefore, when a new queryEmbedding is created from a client request, it would be advisable to have Milvus find the cached vectors most similar to queryEmbedding. Performing the vector similarity search outside of Milvus as is currently done introduces inefficiencies and requires additional testing to verify that the code is correct.