Skill idea: reviewer of python acceleration technology + benchmark python acceleration technology

[mmcky]

I think we should think about two skills that are separate

1. a SKILLS.md only skill that asks a frontier model to review the python acceleration technology we have chosen in the lecture code-cell by code-cell and advise if it is the best choice between (numpy, numba, jax, or another acceleration technology). This is a basic recommender skill that is pretty simple but provides an LLM assessment as an easy go to to start the review process. As models get better this will only improve and keeping skills as simple as possible is more forward looking.

2. a benchmarked -- similar to what we have but simplified to just benchmark and build a table. The skill would take a code-cell and build pure python/numpy vs. numba vs jax implementations (building a temporary version of the unimplemented items, validating they all make the same output and then returning a timing benchmark and validation report).

[mmcky]

@xuanguang-li I think we should think about the skills we have been working on in two stages to simplify. What do you think?

When you have time can you add a comment here about the other dimensions you have in your rubric (other than raw timing) that is taken into consideration in the current benchmark tool.