Improve RAG

[nick863]

Currently we use Vector search for RAG and manually do tokenization and vectorization. I have talked to Pamela Fox, and she proposed the next improvements:

1. Use Hybrid search instead of vector search https://aka.ms/ragrelevance
2. Add semantic ranker to reduce noise from the vector search
3. Use Azure AI search built in vectorization
4. Use Azure AI search built in splitter
5. Change the documentation accordingly.

[plubala]

How does machine learning differ from traditional programming approaches?