Generate smaller enum types

[ian-h-chamberlain]

Closes #166 I guess?

Related to #178 (review) and some of the changes in #162

By passing -fshort-enums to clang, we can get the same smaller types that gcc generates for some of these enums.

This also makes sure the layout tests have the same smaller sizes, so the expected sizes now match the actual sizes of the same C structs.

---

One thing this doesn't do is truly check that GCC is generating the same everything... ideally, I think we'd want to link against a C lib that had a bunch of APIs like this or something:

size_t sizeof_errorType() { return sizeof(errorType); }
size_t alignof_errorType() { return alignof(errorType); }
size_t offsetof_errorType() { return offsetof(errorConf, type); }

and then a Rust test like

extern "C" {
	fn sizeof_errorType() -> libc::size_t;
	fn aligneof_errorType() -> libc::size_t;
	fn offsetof_errorType() -> libc::size_t;
}

#[test]
fn layouts() {
	assert_eq!(mem::size_of<errorType>(), sizeof_errorType());
	assert_eq!(mem::align_of<errorType>(), alignof_errorType());
	assert_eq!(mem::offset_of!(errorConf, type_), offsetof_errorType());
}

But without bindgen doing something like this for us, this seems like a tricky thing to write unless we just do it manually. So for now I think I'll kinda ignore that and leave it as a future enhancement if we really need it.

[Meziu]

Ouch, quite a lot of size changes. I must say it scares me quite a bit to have all of these representations changed on a whim, but it makes sense seeing how there is some inconsistency between the Rust and C repr.

Praying that this won't break some hidden API 🙏 (at least we spare some space compared to before 🤷‍♂️ ).

[FenrirWolf]

If anything the scary/weird part is that gcc seems to use short enums by default for arm-none-eabi while clang doesn't. You'd think they would agree on those settings, but at least we can do it manually this way. We'll probably want a similar PR for citro3d-rs too