Update to pydantic v2

[phackstock]

Closes #291, closes #214.

Changes

Pydantic v2 introduced a lot of API changes and deprecated a lot of classes and functions so quite a few changes were necessary.

CustomErrors

Previously we made use of a lot of custom errors. Pydantic v2 no longer allows this. There is a class PydanticCustomError but it cannot be subclassed.
After consideration I came to the conclusion that it's still useful to make use of these custom errors since this gives us finer control over the formatting of the error messages.
As a fix I ripped out the error folder and replaced it with a single error.py module. In this module there's now a dictionary called pydantic_custom_error_config that defines the values for our various custom errors. For each error this is a short name and the error message.
Custom errors can now be raised like so:

from pydantic_core import PydanticCustomError
from nomenclature.error import custom_pydantic_errors
...
raise PydanticCustomError(*custom_pydantic_errors.MyCustomError, {"value1": value1, "value2": value2})

where value1 and value2 are to be interpolated as part of the error message.

Error collection

Previously I had (ab)-used pydantic achieve the behavior of collecting a number of errors and then raising them all at once. This is no longer possible since pydantic 2 removed the class I used to do that. It was also never really the cleanest way to implement this mechanic since pydantic errors should be reserved for validation at object creation.
I implemented a class called ErrorCollector which collects errors in a list to be raised at a later point in time.
This is how you use it:

    errors = ErrorCollector()
    for file in required_data_dir.iterdir():
        try:
            RequiredDataValidator.from_file(file).validate_with_definition(dsd)
        except ValueError as error:
            errors.append(error)
    if errors:
        raise ValueError(f"Found error(s) in required data files:\n{errors}")

Output wise it is currently not ideal but I'd say this PR is big enough and I'll address this in a follow up PR.

Each item

Pydantic v2 changed the rather straightforward each_item option for validators. Previously this option would tell pydantic to apply a validator to each item of a list of dictionary.
Now this has to be done via the Annotated keyword in the variable type definition.
I don't really like it in terms of readability but there seems to be no other option.

Various small changes

Optional to |

In a number of type hints, updated Optional to the recommended | None.

Used the new model_dump

In a couple of places the new Basemodel.model_dump() came in handy for shortening code.

Removed unused code

• DataStructureConfig.check_repository_consistency was always covered by another validator
• In RegionProcessor._apply_region_processing it is not necessary to check that all the regions are valid. This is already done at initialization of the processor.

[danielhuppmann]

Thank you for the thorough refactoring, looks all good to me!