exclude_none does not exclude_none on non-pydantic models

[matt-mercer]

to_jsonable_python only seems to apply exclude_none for pydantic models, presumbly for dataclass.dataclass or dict types it is just returning asdict or the supplied value for the top level

Expected

exclude_none=True should be honoured for all dict members in result of to_jsonable_python ..
.... I feel that the behaviour should be homogeneous here, since to_jsonable_python is accepting ( and serialising ) none pydantic types, the fact that this top level instruction is ignored, while others are honoured, seems inconsistent.

Test Case

@pydantic.dataclasses.dataclass
class PDCNoneableModel:
    status_code: int = 200
    test1: str = "ok"
    test2: str | None = None

@dataclasses.dataclass
class DataclassNoneable:
    status_code: int = 200
    test1: str = "ok"
    test2: str | None = None

@pytest.mark.parametrize("modellike", [
    PDCNoneableModel(), DataclassNoneable(), {"test1": "ok", "test2": None}
])
async def test_jsonable_exclude_none(modellike: PDCNoneableModel | DataclassNoneable | dict):

    raw_out = to_jsonable_python(modellike, exclude_none=True)

    assert isinstance(raw_out, dict)
    assert raw_out["status_code"] == 200
    assert "test1" in raw_out
    assert "test2" not in raw_out

[matsjoyce-refeyn]

A workaround is to use pydantic.TypeAdapter.dump_*, for which the exclude_none flag does seem to work.

[davidhewitt]

Still reproduces on latest main.