Rename super() to supertype()

[nalimilan]

This name is more explicit and consistent with subtypes().
Also rename promote_to_super() for consistency, but without
deprecation as it isn't exported.

Fixes #14337.

[StefanKarpinski]

I'm in favor of this rename, but I suspect we should wait until after potential shakeup due to adding protocols and/or traits.

[nalimilan]

Why not, but this doesn't strike me as really affecting the same code, nor requiring the same kind of change.

[StefanKarpinski]

Eh, that's fair. I guess I'm just wondering somewhere if the whole notion of super won't go away. But maybe we should just do this anyway. It's a much better name.

[nalimilan]

I'm fine with waiting for a bit until we know whether abstract types might go away completely.

[jakebolewski]

They are not anytime soon.

[JeffreySarnoff]

supertype is an easy preference

and super(x) should be reflexively assignable

[mason-bially]

Also, I think based on my understanding of the proposals, although I might be wrong, but supertype would still be a helpful utility function regardless of the changes (even the proposal technically removing abstract types would still have use for a function named supertype).

[JeffBezanson]

+1
Sorry this took long enough that it needs a rebase. But, let's rebase and merge it.

[nalimilan]

No problem. I've rebased it and will merge when the tests pass.

[tkelman]

please don't restart failed travis jobs without making a note and saving the log.

[nalimilan]

Sorry, I always forget that Travis doesn't keep the log available.