How should the kit work?

[seagreen]

Three options:

1. It's a separate repo, which calls out to bowtie to do everything. At the start all the tests pass, and over time the implementer replaces each call to bowtie with their own implementation of evaluation, type inference, etc.

2. It's a separate repo, with undefined for each major function. At the start the tests fail, and go green as the user implements them

3. It's in the SAME repo (thanks @mitchellwrosen) for this idea. This way things like fuzzy file searching and ghcide jump-to-definition still work. Perhaps it's in the Kit namespace instead of Bowtie.

[seagreen]

In the meantime while I think about this the kit repo is deprecated 🤦‍♂️