BUG: Scipy interpolate methods are not datetime64 aware

[TomAugspurger]

From this SO question.

Is this something we should worry about here, or report upstream to scipy? The workaround I posted on SO is just to explicitly recast the index as an int64, interpolate on that, and then reset the original index. Any reason we shouldn't do this internally?

[jreback]

@TomAugspurger I think you may need to do a ``values.view('i8')` before passing if they are dtype.kind of 'M' or 'm'

reopening for #5977 (comment)