Pandera MultiIndex tz-aware level loses timezone during validation when full materialization path is used

[Owen-OptiGrid]

Describe the bug
When validating a tz‑aware MultiIndex level, Pandera’s full‑materialization path converts the level via .values, which drops timezone info. This leads to a dtype mismatch error: expected datetime64[ns, ] but got
datetime64[ns], even though the MultiIndex level is tz‑aware.

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of pandera. (0.28.1)
• (optional) I have confirmed this bug exists on the main branch of pandera.

Note: Please read this guide (https://matthewrocklin.com/blog/work/2018/02/28/minimal-bug-reports) detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

# /// script
# dependencies = [
#   "pandas>=2.0",
#   "pandera==0.28.1",
# ]
# ///

from __future__ import annotations

from zoneinfo import ZoneInfo

import pandas as pd
import pandera.pandas as pa
from pandera.engines import pandas_engine


def build_data() -> pd.DataFrame:
    tz = ZoneInfo("America/New_York")
    idx = pd.MultiIndex.from_arrays(
        [
            pd.DatetimeIndex(
                ["2024-01-01 00:00", "2024-01-01 01:00"], tz=tz, name="LEVEL_ONE"
            ),
            pd.Index(["A", "B"], name="LEVEL_TWO"),
        ]
    )
    return pd.DataFrame({"value": [1, 2]}, index=idx)


def build_schema(tz: ZoneInfo) -> pa.DataFrameSchema:
    return pa.DataFrameSchema(
        columns={"value": pa.Column(int)},
        index=pa.MultiIndex(
            [
                pa.Index(
                    pandas_engine.DateTime(tz=tz),
                    name="LEVEL_ONE",
                    # Force full-materialization path where Pandera uses .values.
                    checks=pa.Check(lambda s: s.notna()),
                ),
                pa.Index(pa.String, name="LEVEL_TWO"),
            ]
        ),
    )


def main() -> None:
    df = build_data()
    schema = build_schema(ZoneInfo("America/New_York"))
    schema.validate(df)


if __name__ == "__main__":
    main()

Expected behavior

Validation should pass, since the MultiIndex level is tz‑aware and matches the schema’s tz. It should not drop timezone info when validating.

Desktop (please complete the following information):

• OS: Linux (ubuntu/debian family)
• Browser: N/A
• Version: N/A

Screenshots

N/A

Additional context

The failure occurs only when a non‑“unique‑only” check is present, which forces the full‑materialization path. That path calls .values on the tz‑aware level (multiindex.get_level_values(...)), stripping the timezone and
causing a dtype mismatch.