Difference of behavior between versions 0.20.4 and 0.21.0 related to pandas timezone management

[PetitLepton]

Hi fellows,

due to some issues with the management of time zones by pandas, see pydantic/pydantic#10120, we had written a unit test involving pandera as the following

# Standard Library
from datetime import datetime
from zoneinfo import ZoneInfo

# Third-party
import pandas
import pytest

from pandera import DataFrameModel
from pandera.errors import SchemaError
from pandera.typing import DataFrame, Series
from pydantic_core._pydantic_core import TzInfo


def test_bug_pydantic_v2() -> None:
    UTCDateTimeType = pandas.DatetimeTZDtype(tz=ZoneInfo("UTC"))

    class MyDataFrameModel(DataFrameModel):
        timestamp: Series[UTCDateTimeType]

    timestamp = datetime(1970, 1, 1, tzinfo=TzInfo(0))
    with pytest.raises(SchemaError):
        pandas.DataFrame(data={"timestamp": [timestamp]}).pipe(
            DataFrame[MyDataFrameModel]
        )

This test was passing up to version 0.20.4 but fails now with version 0.21.0 (meaning that the schema check is compatible with pydantic). As I am surprised that it is passing, pandas version did not change, I was wondering if the latest modifications in pandera may have hidden such a weird behavior.

Thanks in advance for your input!