How to type return value of `Column` reductions

[MarcoGorelli]

This is related to #155, but I'm opening a new issue to make sure it's not forgotten

Currently column reductions are typed to return type. But if we also use type to type arguments, such as Column.from_sequence, then they should probably be distinguished.

For example, I could pass 'float64' to Column.from_sequence, but the return value of column.mean() wouldn't be 'float64', it would probably just be a Python 'float'.

I don't really know yet what the solution here is, will think about it, but I think we can do better

[rgommers]

(note: it should be dtype not type above)

I meant to revisit that - in dataframe_object.py we're using Scalar as the annotation for Python scalars. E.g.,:

Scalar = TypeVar("Scalar")  # not further worked out, indicates any Python scalar type

def __eq__(self, other: DataFrame | Scalar) -> DataFrame:

Now of course when it's a return type the static typing should technically be overloads I think, rather than dtype or Scalar. But we don't want to clutter the spec itself with overloads probably, because those are pretty meaningless and verbose to the average reader.