Fix `.loc` to handle ambiguity if a single Scalar is first element of a tuple

[Dr-Irv]

From discussion in #575

In

pandas-stubs/pandas-stubs/core/frame.pyi

Line 163 in 41fe0f5

IndexType | MaskType | list[HashableT] | Hashable,

Replace Hashable with slice | _IndexSliceTuple | Callable

If you make the suggested change, that will cause the test test_frame.py:test_loc_slice() to fail, but I now realize that the expression used there is ambiguous:

>>> df1 = pd.DataFrame(
...         {"x": [1, 2, 3, 4]},
...         index=pd.MultiIndex.from_product([[1, 2], ["a", "b"]], names=["num", "let"]),
...     )
>>> df1.loc[1, :]
     x
let
a    1
b    2
>>> df2 = pd.DataFrame({"x": [1,2,3,4]}, index=[10, 20, 30, 40])
>>> df2.loc[10, :]
x    1
Name: 10, dtype: int64

So the first argument as an integer could return a DataFrame or Series, dependent on whether the underlying index is a regular Index or MultiIndex

The solution is then to add another overload in _LocIndexerFrame.__getitem__():

    @overload
    def __getitem__(self, idx: tuple[ScalarT, slice]) -> Series | DataFrame: ...

Then modify the test in test_index_slice() to check that the type is Union[pd.Series, pd.DataFrame], and add another test corresponding to df2 above.

Originally posted by @Dr-Irv in #575 (comment)