Add notebook for generating embeddings and initial RAG implementation for fetching the appropriate layer

[leothomas]

Notes:

• Embeddings for some of the layers from the Confluence document still need to be generated (see note in IPyNb)
• The RAG uses LanceDB as a vector store, backed directly to S3. Appropriate AWS Credentials/IAM role will be required to access the LanceDB instance
• The vector store search step is now implemented as a retriever - I'm not sure if the LangChain model automatically knows to use the retriever of if the prompt needs to be amended to tell the Agent to use the retriever tool first