feat: Compute closure captures

[hkalbasi]

This PR:

• Computes closure captures and the trait it implements (Fn, FnMut or FnOnce)
• Computes data layout of closures
• Adds support for closure MIR lowering
• Changes the closure type display from |arg1: ty1, arg2: ty| -> ret to impl FnX(arg1: ty1, arg2: ty2) -> ret

fix #12297

[hkalbasi]

Is there a way in salsa to get a query result only if it is calculated before? Currently, the display code for closure calls db.infer, which works, but if we try to use it in the middle of infer (for example for print debugging, or due some future bug) we will get cycle panic, which is worse than fallback to a simpler display.

[bors]

☔ The latest upstream changes (presumably #14436) made this pull request unmergeable. Please resolve the merge conflicts.

[Veykril]

Haven't looked over everything yet, but as this is a rather big PR again, could you try splitting out the derive macro expansion change as its own PR? And if possible also the Forces reborrow of mutable references in method receiver, part

[bors]

☔ The latest upstream changes (presumably #14509) made this pull request unmergeable. Please resolve the merge conflicts.

[lowr]

I think #12297 should be linked?

I don't have too strong opinion here, but I'd rather have inlay hints rendering for closures configurable. I pretty much agree with this comment in that I don't usually care what Fn* traits closures implement. Current pseudo-Rust notation |T, U| -> R is not valid Rust, but I find it concise which I value for stuffs shown in my editor all the time.

On the other hand, I'd love to see the information on hover!

[hkalbasi]

Ah, I forgot that it was controversial last time. Will make it configurable.

[hkalbasi]

@bors r+

[bors]

📌 Commit 59b6f2d has been approved by HKalbasi

It is now in the queue for this repository.

[bors]

⌛ Testing commit 59b6f2d with merge 44cf8ef...

[bors]

☀️ Test successful - checks-actions
Approved by: HKalbasi
Pushing 44cf8ef to master...

[Veykril]

This absolutely ruined type inference on self (unknown types)

[hkalbasi]

Ah sorry, I should check these always. Will look at it.

[Veykril]

Ah I think this is in part due to the dangling temporary Expr::Missing we are allocating now, though we do have some changes to the pattern types as well

[hkalbasi]

Type inference for nested closures is broken it seems. Still looking.

[Veykril]

Should we revert this PR for now (so you don't have the stress of having to fix this immediately?), and maybe just pick the hir-def pieces and what not that don't break things yet

[hkalbasi]

Since this is big, and I have some uncommited changes on top of it, reverting it makes some git headache for me. I will file a revert tomorrow if the fix is not trivial.

[hkalbasi]

@Veykril it looks like metric is now back, but I didn't change anything related to the dangling Missing. Was it just a guess, or did you see something about it?

[Veykril]

Just an assumption thoug I mightve misunderstood parts of the code. Do note thoug that we are not fully back to prior this PR in terms on unknown types, so it would be nice to figure out what else regressed here.