feat: Automatic evaluation of RAG pipeline

[KevinJBoyer]

Ticket

As part of https://navalabs.atlassian.net/browse/DST-180, we want to first provide a way to automatically evaluate the impact of various changes to the RAG pipeline on accuracy.

Changes

• Add eval.py, which uses the Phoenix evals prompt to compare human ground truth with AI-generated answers and logs the results to a .csv file
• Update ingest.py to accept chunk_size and overlap_size as parameters, and provide a silent parameter so they don't print to console

Context for reviewers

• This is a manual version to get a feel for what it's like to implement this directly -- I plan to try out an existing framework (like Phoenix) next
• Preliminary results are available at https://docs.google.com/spreadsheets/d/1ZAZGZsqMGg2dX5qdy0j_n0xEAigWvffMPqvzvhD1DPk/edit#gid=687224843

Testing

• python eval.py. If you use one of the OpenAI LLMs (either for generating answers or evaluating them), set OPENAI_API_KEY in your environment
• You can modify the parameters that are tested by updating parameters -- but note that it can quickly get expensive. Trying 18 total combinations and using GPT-4 Turbo to generate half of the answers + evaluate all of the answers cost $3.18.

[yoomlam]

Nice, clean code!