Handle UnivCo

[conal]

Handle UnivCo: navigation, transformation, pretty-printing. I was in some unfamiliar territory, and I don't have any tests. Some questions:

• ppCoercion in HERMIT.PrettyPrinter.AST: should there be a catch-all here as in ppCoercionR from HERMIT.PrettyPrinter.Clean?
• I'm skipping UnivCo in allR in the Walker c Coercion instance in HERMIT.Kure. Is that what we want? I guess so, as with Refl.
• In HERMIT.PrettyPrinter.Clean.ppCoercion, for the UnivCo case, should I use retApps?

[conal]

I see the "Merge remote-tracking branch 'upstream/master'" commit listed here (presumably for @xich's merge of my previous pull request) in addition to my new content. Am I doing something wonky with git?

[conal]

See conversation at #160.

[xich]

I think, instead of a catch-all case, we should use readerT (http://hackage.haskell.org/package/kure-2.16.12/docs/Language-KURE-Combinators-Transform.html) so we get an exhaustiveness check from GHC... but we can do that in another diff.

Yes, no need for UnivCo in the Walker c Coercion instance because a UnivCo doesn't contain other coercions.

Yes, I would use retApps so the types get parens around them if necessary.

I think the merge commit would disappear if you rebased this onto my merge commit for SubCo, but that would probably require a git push -f... we can just see what happens to it when I merge this in. Mind doing the retApps bit? Otherwise looks good!

[conal]

@xich Thanks. I added a commit with retApps.

Makes sense about UnivCo having no sub-conversions.

I need to learn git better.

[xich]

Eh, git I understand... github, on the other hand, is still a mystery much of the time.

Thanks for the patches!