Some tooling experiments

[ianthomas23]

This is trying out a few experiments on tooling, and adding a few more example schemas. It is based on top of #1 so ideally that would be merged and this rebased on top of main.

There is a kernel spec schema based on jupyter/enhancement-proposals#105 and a couple of kernel message schemas so that we have one schema reusing another, but I don't think I have the $ref between the two working yet. The schema names follow that outlined in #2.

There is some pytest testing that validates each of the schemas against the meta schema (2020-12 Draft) and also checks that the directory and schema names are consistent and so on. There is a simple sphinx build (pip install .; cd docs; make html) of a single schema showing it rendered using sphinx-jsonschema and also in JSON, TOML and YAML formats. I we are going to allow submission of new schemas in those formats then we need to be able to display existing schemas in those formats to inform the user. The contents are hardcoded now but would be dynamically generated for each of the schemas in the repo. The sphinx-jsonschema output isn't excellent (it omits if-then clauses for example). TOML is also problematic as the dump functionality in toml reorders the elements so that the if and when are not longer in the correct order, so we would have to override the output-ordering of the toml writer.

schema.webm

Some of this is different to that proposed in #3 but 🤷.

[Zsailer]

I wrote a very similar class to this in Jupyter Events here.

I wonder if we can consolidate some of this logic.