add ci

[vwoloszyn]

Goal:
The primary objective of this pull request is to implement Continuous Integration (CI) tests, ensuring that new pull requests (PRs) do not adversely impact the ongoing development process.

Why This Matters:
By incorporating CI tests into the development workflow, we establish a proactive approach to identify and address issues early in the PR lifecycle. This not only enhances the overall reliability of the codebase but also streamlines the review process for future contributions.

[jart]

The value of having a good CI needs no explanation. I can't thank you enough for doing this for me! I'm very happy to commit to maintaining any non-flaky CI jobs you and others contribute! For example, it might be nice to have Linux CI too (which is less flaky than MacOS VMs from GA in my experience). I'll try to figure that out at some point. But if you want to take a crack at adding it, you'll likely need a binfmt_misc interpreter setup similar to how I'm doing it here: https://github.com/jart/cosmopolitan/blob/master/.github/workflows/build.yml

[jart]

Looks like the CI run completed in 5 minutes for Mistral. Not bad! Some next steps I'd like to take are:

1. There are examples of tinier weights we could potentially check-in from the upstream project here: https://github.com/ggerganov/llama.cpp/tree/master/models Those could be useful for testing.
2. It might be nice to set --temp 0 or perhaps pick a random seed and verify deterministic output of some random prompt, just to sanity check the implementation, and check that we haven't done anything to break the math.