feat: FAISS in OpenSearch: Support HNSW for cosine

[tstadel]

Related Issues

• closes #2913

Proposed Changes:

• normalize vectors manually at index time
• normalize vectors manually at query time
• use dot_product space under the hood
• make normalization more efficient by using two-dimensional numpy array instead of normalizing each vector separately
• run cosine tests for all document stores
• run all document_store integration tests (not only OpenSearch)
• fix test_faiss_and_milvus.py tests not running

pushed to another PR:

• streamline validation of embedding shape across all document stores (BaseDocumentStore_validate_embeddings_shape)
• introduce DenseRetriever abstraction to facilitate usage of retrievers in document store's update_embeddings

How did you test it?

• made test test_cosine_similarity of test_faiss_and_milvus.py generic to run on all document stores

Notes for the reviewer

Checklist

• [x] I have read the contributors guidelines and the code of conduct
• [ ] I have updated the related issue with new insights and changes
• [x] I added tests that demonstrate the correct behavior of the change
• [x] I've used the conventional commit convention for my PR title
• [x] I documented my code
• [x] I ran pre-commit hooks and fixed any issue

[review-notebook-app[bot]]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB