[spike]: training AI agents on Aztec

[critesjosh]

AI agents could great assist us in teaching developers about how Aztec works. We should do some research into the available options and pick one to explore in more depth.

Things to consider when evaluating:

• easy to train. We would likely want to retrain it every week or so, based on docs and source code updates
• it would be nice if we could train it on the monorepo
• must publicly accessible once trained
• integrating with docs would be nice

Output:

• a document that reviews some AI options, with a recommendation for an AI to use, or a suggestion for when we should re-evaluate