Handle required arguments in nested pydantic models

[Xallt]

from pydantic import BaseModel, Field
import tyro

class A(BaseModel):
    x: str

class B(BaseModel):
    a: A = Field(default_factory=A)

if __name__ == "__main__":
    tyro.cli(B)

This particular example failed on me with

Traceback (most recent call last):
  File "/home/xallt/progs/test.py", line 11, in <module>
    tyro.cli(B)
  File "/home/xallt/clones/tyro/src/tyro/_cli.py", line 207, in cli
    output = _cli_impl(
  File "/home/xallt/clones/tyro/src/tyro/_cli.py", line 332, in _cli_impl
    if not _fields.is_nested_type(cast(type, f), default_instance_internal):
  File "/home/xallt/clones/tyro/src/tyro/_unsafe_cache.py", line 33, in wrapped_f
    out = f(*args, **kwargs)
  File "/home/xallt/clones/tyro/src/tyro/_fields.py", line 248, in is_nested_type
    _try_field_list_from_callable(typ, default_instance),
  File "/home/xallt/clones/tyro/src/tyro/_fields.py", line 417, in _try_field_list_from_callable
    return _field_list_from_pydantic(cls, default_instance)
  File "/home/xallt/clones/tyro/src/tyro/_fields.py", line 689, in _field_list_from_pydantic
    default, is_default_from_default_instance = _get_pydantic_v2_field_default(
  File "/home/xallt/clones/tyro/src/tyro/_fields.py", line 1186, in _get_pydantic_v2_field_default
    return field.get_default(call_default_factory=True), False
  File "/media/xallt/HardDrive/.virtualenv/python310/lib/python3.10/site-packages/pydantic/fields.py", line 556, in get_default
    return self.default_factory()
  File "/media/xallt/HardDrive/.virtualenv/python310/lib/python3.10/site-packages/pydantic/main.py", line 193, in __init__
    self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 1 validation error for A
x
  Field required [type=missing, input_value={}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.8/v/missing

I dug around in the code and realized that this is because of the way that the default values for pydantic fields are retrieved in https://github.com/brentyi/tyro/blob/main/src/tyro/_fields.py#L1186
By wrapping in a ValidationError try-catch, we can check if the specified field did try to initialize but crashed due to lack of constructors for some fields, therefore it is in fact a "required" field

[brentyi]

Hi @Xallt, thanks for the PR!

To clarify, isn't this use of default_factory= invalid?

class B(BaseModel):
    a: A = Field(default_factory=A)

default_factory= should take an instance of Callable[[], Any], but here A would expect an argument.