chore: cleanup for 3.11

[henryiii]

Some cleanup, focused on some things 3.11 provides.

• chore: cleanup for 3.11+
• chore: use StrEnum from Python 3.11
• chore: some cleanup from ruff --preview checks

[joerick]

I had a thought - can we remove from __future__ import annotations now? I can't remember how all that shook out but I think annotations are deferred by default now? Maybe worth trying it and seeing what breaks!

[henryiii]

I think you are referring to Python π making them lazy by default? However, we might not need the annotations because the main reason it's used is to enable invalid expressions that became valid in 3.9 and 3.10 (and stuff that is often better spelled as Self). So I could try and see.

String annotations are still slightly faster, but not measurably for something like cibuildwheel. ;)

[joerick]

Ahh, yeah it's the expression thing. So we might still want it to use more modern type features than exist in 3.11. So I guess 3.14 is where it's gonna get simple again :)

[henryiii]

I removed from __future__ import annotations. A few places needed strings, lots of things I moved to Self, and I was able to remove the one place we were using TYPE_CHECKING since the thing it was protecting was valid on 3.10+. A few things needed reordering since the class must exist before it's in an annotation.

I also added the Path checks, and fixed all of them except the open -> Path.open one, which I ignored for now, as it would require more changes.

[henryiii]

(PS: I can split this up, commits are kept focused on one-ish thing. The future import one is large enough to be a PR if you'd rather)

[joerick]

Looks good to me.

[henryiii]

FYI, I tried Self on these (with @classmethod), but typing didn't seem to understand that enum is final (even if you tell it it's final), so it sees cls.wasm32 as Architecture.wasm32 and therefore not Self. Wrapping it in an extra cls() does make it happy, but cls(cls.wasm32) just to make the type checker happy wasn't worth it. Just added strings here instead of Self. It's technically correct since enums are always final.

[henryiii]

This is a fix, I believe.