Standardize RAG evaluation pipelines

[dgarijo]

Right now some are in python, some in notebooks...
It would be great to use this: https://docs.ragas.io/en/stable/getstarted/index.html to make sure we have the experiments consistently defined. Ideally the code should be reusable between models

It looks like we may be able to reuse the correctness of the answers. Or maybe we can add our own metrics if needed https://docs.ragas.io/en/stable/concepts/metrics/answer_correctness.html