CI: Numpydev failing

[TomAugspurger]

I can reproduce with numpy master

==================================== ERRORS ====================================
__________ ERROR collecting pandas/tests/arrays/sparse/test_array.py ___________
pandas/tests/arrays/sparse/test_array.py:22: in <module>
    class TestSparseArray:
pandas/tests/arrays/sparse/test_array.py:511: in TestSparseArray
    dtype=SparseDtype("datetime64[ns]", pd.Timestamp("1970")),
pandas/core/arrays/sparse/array.py:364: in __init__
    data, kind=kind, fill_value=fill_value, dtype=dtype
pandas/core/arrays/sparse/array.py:1584: in make_sparse
    sparsified_values = arr[mask]
E   IndexError: arrays used as indices must be of integer (or boolean) type
--------- generated xml file: /home/vsts/work/1/s/test-data-single.xml ---------
========================== slowest 10 test durations ===========================

(Pdb) pp arr
array(['1970-01-01T00:00:00.000000000', '1970-01-01T00:00:00.000000001'],
      dtype='datetime64[ns]')
(Pdb) pp fill_value
Timestamp('1970-01-01 00:00:00')
(Pdb) pp arr != fill_value
array([True, True], dtype=object)

Previously, that was bool dtype

In [6]: np.array(['1970-01-01T00:00:00.000000000', '1970-01-01T00:00:00.000000001'],
   ...:       dtype='datetime64[ns]') == pd.Timestamp('1970-01-01 00:00:00')
Out[6]: array([False, False])

[TomAugspurger]

See numpy/numpy#14800 (comment) for whats going on. Not sure if it's a regression or deliberate. Will work around it for now in #29433