Refactor Python code

[evetion]

Fixes #512

• Add JSON Schema of Config and generate Python code
• Autogenerate Pandera Schemas
• Update docs to include the above
• Refactor Python model into split Config & Root submodels

[evetion]

I started with splitting the Model into separate tables and config, but didn't get far without hitting the limits of the current design.

The problem is, there's no clear design, it's mix of different abstractions. Pandera DataFrames are high-level abstract things, but represent low-level tables (which can come from arrow or geopackage). There's a high-level abstraction of a nodetype in between, but it doesn't represent something low-level. The current Model is a mix of a representation of these nodetypes with a (partial) low-level config. Will make a separate issue about this, which needs a refinement.

[Hofer-Julian]

A few, first comments.
I also see that mypy is still unhappy with you.

Will documentation still be part of this PR?

[Hofer-Julian]

Why is pandera optional?

[evetion]

Because I hacked the remarks field into each type specifically for pandera. But the generation of the Config doesn't need it.

Frankly, this whole file is a hack, but it works. Doing it right requires a new package that actually implements the whole of JSONSchema.

[evetion]

I also see that mypy is still unhappy with you.

Yeah, not sure if I can fix that nicely, I dynamically add things to a module namespace. Will try though.

Will documentation still be part of this PR?

Yep, at least about the generation of the config.py files.

[visr]

Is it possible to remove the timestamp from the generated python files?

It creates merge conflicts:

<<<<<<< HEAD
#   timestamp: 2023-09-05T14:00:34+00:00
||||||| 1ad1340
#   timestamp: 2023-08-21T14:05:19+00:00
=======
#   timestamp: 2023-09-04T11:45:02+00:00
>>>>>>> main

And with #594 I expect we will run it more often as part of tasks, even if it is not strictly needed. In that case all it would do is update the timestamp.