Extend Scalar sub-type of Index to the __iter__() method

[gandhis1]

• Closes DataFrame columns should return a type containing Hashable (or Scalar?) elements #365
• Tests added: Please use assert_type() to assert the type of any return value

As mentioned in that issue, I am a little bit confused about what the type of a column or index label is.....feels like in some places we decided it is a Hashable and in other places we decided it's a Scalar. But just to stay consistent with the way this was already annotated, I continued to use Scalar.

[gandhis1]

x = pd.DataFrame([[1, 2, 3], [4, 5, 6], [7, 8, 9]], index=["a", "b", "c"])

   0  1  2
a  1  2  3
b  4  5  6
c  7  8  9

x[[0, 1]] works

x[(0, 1)] does not, because that is interpreted as a column label

[gandhis1]

So it appears NDArray is being treated as a Hashable, which creates overlapping overloads per MyPy. I don't see any way to fix this currently.

So I took a different approach. Went back to annotating using Scalar, but then augmenting that Union with a Tuple[Scalar, ...]. I added some tests to illustrate some of the edge cases. There is one old test that is failing, and not sure what the best course of action there is, I added a note.

[Dr-Irv]

thanks @gandhis1