[vertexai] Add Hybrid Search Capabilities to VertexAI Vector Search

[lspataroG]

🆕 New Feature: Add Hybrid Search Capability to Vector Search, see here.

• Introduced function add_texts_with_embeddings to be able to pass both dense and sparse embeddings (useful anyway as many users already have precomputed embeddings).
• Modified function similarity_search_by_vector_with_score to take in input a sparse embedding in addition to the dense one, added parameter rrf_ranking_alpha (default to 1) to be able to weight between dense and sparse importance during search.
• If a sparse embedding is passed then similarity_search_by_vector_with_score return both scores inside a dictionary with format: [(Document, {"sparse_score": 0.2, "dense_score": 0.7})], otherwise the previous output format is preserved: [(Document, 0.7)].
• Add tests: test_add_texts_with_embeddings and test_similarity_search_by_vector_with_score_output_shape

[lspataroG]

@lkuligin I am not sure why it is timing out