Is your feature request related to a problem? Please describe.

One of the reasons why I want to use pydantic instead of json-schema (where the source of truth for data models I'm using lie) is because if much easier error codes and information when data doesn't validate. However, when translating a json schema that specifies that one of or any of a set of requirements must be met, it creates a bunch of separate data models and puts them together in a single data model that is a union. This results in error codes that are....not very helpful. It creates a laundry list of all the possible sub models that aren't being met.

...
"oneOf": [
   {"required":["option_1"]},
   {"required":["option_2"]},
   {"required":["option_3"]},
...

class ParentModel1(BaseModel):
    other_info:....
    option_1: Option1
    option_2: Optional[Option2] = None
    option_3: Optional[Option3] = None

class ParentModel2(BaseModel):
    other_info:....
    option_1: Optional[Option1] = None
    option_2: Option2
    option_3: Optional[Option3] = None

class ParentModel3(BaseModel):
    other_info:....
    option_1: Optional[Option1] = None
    option_2: Optional[Option2] = None
    option_3: Option3

class ParentModel(
    RootModel[
        Union[
            ParentModel1,
            ParentModel2,
            ParentModel3, #note that in my case I have a few dozen of these...
        ]
    ]
):
    root: Annotated[
        Union[
            ParentModel1,
            ParentModel2,
            ParentModel3,
        ],
        Field(title='Parent Schema'),
    ]

Describe the solution you'd like
Translation of a "oneOf" statement (and similar) to a pydantic @rootmodel validator.

Describe alternatives you've considered
After spending a few hours trying to accurately post-process the resulting data models, I currently edit the resulting data models by hand, which isn't traceable or reproducible in a data pipeline.

I also did a search to see if other issues like this had been opened and couldn't immediately find any - which was surprising to me so maybe I missed something?

Additional context
Add any other context or screenshots about the feature request here.