[Transition PyPIConGPU to pydantic] Proof-of-concept II: Remove schemata

[chillenzer]

• Rebase after merging [Transition PyPIConGPU to pydantic] Proof-of-concept I: Remove _get_serialized #5501. Until the rebase, only the last three commits are new content.

This is the second of a series of PRs relating to #5500.

This introduces one new functionality:
The schema validation falls back to requesting a schema directly from the class (which is hopefully a pydantic.BaseModel in that case) if it can't find a pertinent one in the schema registry. This unfortunately only works for the immediate class at hand and does not apply to following references in other schemata. The latter would involve some bookkeeping to map the referenced URI to the corresponding Python class. Playing around with that a little yielded no really satisfying solutions.

Given that

• all of this is transitional technology that's supposed to be removed after the transition is completed
• and Pydantic already handles the validation for us and is supposed to become the ground truth eventually

I haven't put to much effort into making referencing work. Instead, we can put wildcards where there are references currently.

I've applied this to two examples from #5501: Grid3D and TimeStepSpec, but not Auto. Removing the reference in plugin.Plugin.json would allow for plugins to contain arbitrary content and effectively circumvent all schema validation which is not what we want. Removing the plugin schemata should consider all plugins in one go instead.

[chillenzer]

Ready for review but the CI is not up to the task yet.
Still, feel free to have a look @pordyna and @PrometheusPi!

[pordyna]

Looks good to me! Out of curiosity, how does such a pydantic generated schema look like?

[pordyna]

@PrometheusPi Do you still want to take a look? Otherwise, you can just merge it.

[chillenzer]

Looks good to me! Out of curiosity, how does such a pydantic generated schema look like?

Well, let's try:

>>> from pydantic import BaseModel
>>> class A(BaseModel):
...     a: str='asdf'      
...     b: int=42
... 
>>> A.model_json_schema()
{'properties': {'a': {'default': 'asdf', 'title': 'A', 'type': 'string'}, 'b': {'default': 42, 'title': 'B', 'type': 'integer'}}, 'title': 'A', 'type': 'object'}