This changes the behavior for:
* The `deg`  and `axis` arguments of `<type>der`
* The `deg`  and `axis` arguments of `<type>int`
* The `deg`  argument of `<type>gauss`
* The `deg`  argument of `<type>vander2d`
* The `deg`  argument of `<type>vander3d`

The old behavior was:
* Raise `ValueError` if the argument is a float, but not an integral one
* Allow a float like `1.0` to mean `1`.

This is inconsistent with most other integer-accepting APIs in numpy, which require these to be actual integers, and raise TypeError when they are not.

The new behavior is:
* Raise `TypeError` if the argument is a float, but not an integral one
* Emit a `DeprecationWarning` if a float like `1.0` is passed, continuing to allow it its old meaning.

This fix will preserve subtypes of ndarray when given as input (x)
to the polyval function. In particular, the results for masked
values of a masked array will be masked.

Fixes #2477.

Choices made, and the alternatives rejected (for no particularly strong reason):

1. Show terms in ascending order, to match their internal representation
  * alternative: descending, to match convention
2. Shows 0 terms in gray
  * alternative: omit entirely
  * alternative: show normally to aid comparison
3. Write each term as `basis(ax + b)
  * alternative: write as `basis(u) ... where u = ax + b`
  * alternative: show the normalized polynomial

In future it would perhaps make sense to expose these options to the end user