TST: use custom parametrization for consistency in base extension array tests

[jorisvandenbossche]

Follow-up on #44242, as that broke geopandas CI. This adds back the custom parametrization, which also makes it consistent with the other occurrence of this test (for a different op) which is still using this:

pandas/pandas/tests/extension/base/ops.py

Lines 116 to 127 in 057c6f8

	@pytest.mark.parametrize("box", [pd.Series, pd.DataFrame])
	def test_direct_arith_with_ndframe_returns_not_implemented(self, data, box):
	# EAs should return NotImplemented for ops with Series/DataFrame
	# Pandas takes care of unboxing the series and calling the EA's op.
	other = pd.Series(data)
	if box is pd.DataFrame:
	other = other.to_frame()
	if hasattr(data, "__add__"):
	result = data.__add__(other)
	assert result is NotImplemented
	else:
	raise pytest.skip(f"{type(data).__name__} does not implement add")

Since this is the only occurrence of frame_or_series in the base extension tests, I think it's not warranted to make it a fixture. If we would rather want this, we should add it to tests/extension/conftest.py instead (can always be done in a follow-up).

[jbrockmendel]

we should add it to tests/extension/conftest.py

i dont understand. why would we duplicate the fixture already in the top-level conftest?

[jorisvandenbossche]

why would we duplicate the fixture already in the top-level conftest?

Because it makes it clear what is expected to be implemented by the downstream user. In practice there are currently no duplicated fixtures though.

[jorisvandenbossche]

Since this will fixup downstream CI + make this consistent within the same file, going to merge this. If we want to change the use of the fixture, we can do that in a later PR.