Use pydantic v2 models instead of v1 models in spacy/schemas.py

[NixBiks]

Currently in schemas.py pydantic imports are done the following way

try:
    from pydantic.v1 import (
        BaseModel,
        ConstrainedStr,
        Field,
        StrictBool,
        StrictFloat,
        StrictInt,
        StrictStr,
        ValidationError,
        create_model,
        validator,
    )
    from pydantic.v1.main import ModelMetaclass
except ImportError:
    from pydantic import (  # type: ignore
        BaseModel,
        ConstrainedStr,
        Field,
        StrictBool,
        StrictFloat,
        StrictInt,
        StrictStr,
        ValidationError,
        create_model,
        validator,
    )
    from pydantic.main import ModelMetaclass  # type: ignore

This makes sure that spacy supports both v1 and v2 of pydantic. The downside is that only v1 models are used which means you can't extend any of the models in v2 models. It also means that performance will be worse.

I have a use case where I need to wrap TokenPattern in a pydantic v2 model which isn't possible. Instead I have to redo all the models in schemas.py as v2 models but maybe you want that change as a PR instead? I realise it might not be straightforward to use v2 models when possible and v1 models otherwise but I think it's worth it - even just for the improved performance. I imagine a solution will involve schemas/__init__.py, schemas/pydantic_v1.py and schemas/pydantic_v2.py.

I guess it's also worth considering dropping v1 support. Let me know what you think, thanks.