Docsite search: Add a new docsite search RAG

[hanna-paasivirta]

Add a new docsite search RAG to replace the experimental Search service. Use the Search service as an example, but rewrite to match the structure of our new embeddings services.

• Leverage the Embeddings service and its SearchResult and VectorStore classes and add module to connect to docs embeddings
• Add separate service to embed docs
• Replace the Search service (doesn't seem to be in use) with a new docsite search service that can search by connecting to the Embeddings service
• Use the same vector store and embeddings APIs as new services (Pinecone, OpenAI)
• Analyse and optimise the chunking of the docsite and adaptor APIs
• Include metadata for each document returned to present to the user

[josephjclark]

Can confirm that the existing search is not in use and we can freely rename it.

To consider (and let's call Elias): is the VectorStore abstraction helping at all or should we just use langchain directly?

Update: let's drop vector store and just use langchain

Structure:

embeddings/
    -docs_store.py <-- this will 
docs_search.py   <-- this is a connected service which will search the docsite and return useful document chunks (it might be very lightweight)