Write dataset description for all converters

[ghisvail]

Dataset description files are a requirement for BIDS compliance.

• Define a proper model for the BIDS dataset description
• Ensure the model is used in bids utility functions
• Add missing dataset description for some converters

[ghisvail]

Note that the current definition of the BIDS dataset description model is the most nominal one based on the current specs. But it's implemented in a way that future extension only requires adding a properly typed attribute to the class and a field-to-alias mapping in Config.fields if the field is not serialized to pascal case.

[ghisvail]

Blocked by #392, which implements a dataset description model for HABS that could be used for the other converters targeted in this PR. Depends which one we want to merge first between #392 and this one.

[ghisvail]

We still need to update the CI data to include the missing dataset_description.json file for some converters. To be done after the current CI refactoring and testing efforts.

[ghisvail]

Before we consider merging this, I'd like to test the same approach with attrs (an alternative to pydantic), which is used by pydra extensively. If we were to consider switching to pydra at some point, we might benefit from using the same dependencies.

[ghisvail]

I believe the alternative solution in #535 using attrs is better from a conceptual and maintenance perspective. No need to learn the intricacies of pydantic (especially the Config inner class) and closer to the composition over inheritence principle where Python has been going lately. The attrs solution is basically a dataclass (record type) composed with a converter registry provided by cattrs. Simple.

[omar-rifai]

Should we close this PR?

[ghisvail]

It will be closed automatically once #535 is merged