Add FnDef to TypeName

[nathanwhit]

Addresses the FnDef part of #368.

Lowering and program clause generation are probably the areas that need the most review. I'm also a bit unsure about if/how the builtin traits apply to FnDef, so some guidance on that would be helpful. For parsing I've chosen to only handle function declarations for simplicity's sake (plus I'm not sure why we would need to inspect the function bodies).

In its current state the PR could use some refactoring/polish, but I'll defer that until the more substantive aspects of the PR are in good shape.

[nikomatsakis]

This looks really good! The main thing I think we might want is to try and cleanup the program clause generation code, but maybe that's better left for another PR.

(I'm also pondering the implied bounds generation logic, but I think it's probably what we want, even though it can't actually be observed by users at the moment.)

[nathanwhit]

Thanks for the helpful feedback!

I've added the implied bounds test and cleaned up the program clause generation code quite a bit, so it should be ready for re-review.

[nikomatsakis]

Requested a few nits! Looking very close to me.

[nikomatsakis]

I feel like we are going to need .. well, eventually we'll have to add ABI and other details, but maybe we can add that together with adding the same details to the function pointer types

[nikomatsakis]

r=me but this needs to be rebased.

[nikomatsakis]

And we should probably file a followup issue regarding ABI details (which apply equally to function pointers)

[nikomatsakis]

r=me but requires rebase and we should break out a follow-up issue regarding handling ABIs, both in these types and function pointers.

[nathanwhit]

Done and done!