[CI][Python] Dask integration nightly jobs fail on dask.dataframe.io.tests.test_parquet::test_pandas_timestamp_overflow_pyarrow

[raulcd]

Describe the bug, including details regarding any error messages, version, and platform.

Dask nightlies have been failing since: #33321 was merged.

=================================== FAILURES ===================================
____________________ test_pandas_timestamp_overflow_pyarrow ____________________

tmpdir = local('/tmp/pytest-of-root/pytest-1/test_pandas_timestamp_overflow0')

    @PYARROW_MARK
    def test_pandas_timestamp_overflow_pyarrow(tmpdir):
        info = np.iinfo(np.dtype("int64"))
        # In `numpy=1.24.0` NumPy warns when an overflow is encountered when casting from float to int
        # https://numpy.org/doc/stable/release/1.24.0-notes.html#numpy-now-gives-floating-point-errors-in-casts
        if _numpy_124:
            ctx = pytest.warns(RuntimeWarning, match="invalid value encountered in cast")
        else:
            ctx = contextlib.nullcontext()
        with ctx:
            arr_numeric = np.linspace(
                start=info.min + 2, stop=info.max, num=1024, dtype="int64"
            )
        arr_dates = arr_numeric.astype("datetime64[ms]")
    
        table = pa.Table.from_arrays([pa.array(arr_dates)], names=["ts"])
        pa.parquet.write_table(
            table, f"{tmpdir}/file.parquet", use_deprecated_int96_timestamps=False
        )
    
        # This will raise by default due to overflow
>       with pytest.raises(pa.lib.ArrowInvalid) as e:
E       Failed: DID NOT RAISE <class 'pyarrow.lib.ArrowInvalid'>

Component(s)

Continuous Integration, Python