Unify RAG Implementation

[tmichaeldb]

Currently, we have several modules that implement RAG in some way:

• rag_handler.py
• rag.py
• knowledge_base/controller.py
• email_rag repository

Additionally, our RAG implementation in the mindsdb main repo is outdated in comparison to the email_rag repository. Therefore, we should take this as an opportunity to bring our learning back into our main repository, and make the RAG implementation as clean as possible.

Specifically, we need to:

• Make sure our RAG implementation aligns with our Agents Roadmap
• Combine current implementations into one fully functional interface for RAG and deprecate/rework others (e.g. all RAG is through Agents + Knowledge Bases)
• Bring RAG improvements from email_rag back into our main repo (e.g. better performing retrievers, better chunking techniques, etc)