Ignore Pydantic dataclasses

[charles-dyfis-net]

Description

While Pydantic dataclasses have a __dataclass_fields__ attribute, their users may have validation logic, default_factory configuration, and other settings that msgspec does not enforce.

Rather than appearing to support these types but silently ignoring validation steps and defaults, it makes more sense to treat them as explicitly unsupported.

[timinou]

Allow me to drop in :)

I'm writing a framework that unfortunately has to handle a few libraries that use Pydantic. I think having a basic msgspec <> Pydantic model translation would help me leverage the outputs of those libraries without having to create a second source of truth within my code. Even if PR #729 gets accepted, a semi-blessed match-case statement here with msgspec inspect's type would go a long way in ensuring some type of compatibility.

Thanks!

[emosenkis]

I would also like to use msgspec together with some external types that use pydantic. Rather than making msgspec always fail in that case, it would be much nicer to allow users to opt in to a compatibility layer (which could be as simple as delegating conversion to and from simple types to pydantic). It's better to offer a lower performance path than to break anyone that might be using this just in case they might be incorrectly expecting validations to be enforced.