Implement Deref<Target = [T]> for [T; N], and fix const infer variable canonicalization (fix Deref<Target=[T; N]> ICEs)

[compiler-errors]

1. Stash a const infer's type into the canonical var during canonicalization, so we can recreate the fresh const infer with that same type.
   For example, given [T; _] we know _ is a usize. If we go from infer => canonical => infer, we shouldn't forget that variable is a usize.
   Fixes Inferring const parameters to wrapper type causes rustc to panic. #92626
   Fixes ICE on Rust 1.51 with min const generics and Deref<Target=[T; N]> #83704

2. Implement Deref for [T; N] which allows us to remove some custom array unsizing logic that was leaking an inference variable during method selection.
   Fixes Calling slice method on newtype wrapper that implements Deref<Target=[_; 1]> causes ICE #92637

r? @lcnr

[compiler-errors]

This is the only part of the PR I am somewhat unhappy with, but it works.

[eddyb]

Trait impls are insta-stable and I strongly disagree with a [T; N] -> [T] deref impl, regardless.

(click to expand my reasons - they're not really relevant outside of a FCP/RFC discussion)

The reason deref impls like Vec<T> -> [T] exist is because Vec<T> is like Box<[T]>, i.e. a pointer to [T].

Array unsizing (to slice) is in a different coercion category than deref coercions, and more flexible at that (e.g. Rc<[T; N]> -> Rc<[T]>, where a deref coercion couldn't give back an Rc) - I also don't see any reason to mix up the two too much (though long-term I would like to see generalized coercions).

Something this significant requires at least an FCP by @rust-lang/lang, if not an RFC.

[rust-log-analyzer]

The job x86_64-gnu-llvm-12 failed! Check out the build log: (web) (plain)

Click to see the possible cause of the failure (guessed by this bot)

Set({"src/doc/edition-guide"}) not skipped for "bootstrap::doc::EditionGuide" -- not in ["src/tools/tidy"]
Rustbook (x86_64-unknown-linux-gnu) - edition-guide
Building stage0 tool linkchecker (x86_64-unknown-linux-gnu)
    Finished release [optimized] target(s) in 0.16s
core/primitive.array.html:336: broken link fragment `#method.to_ascii_uppercase` pointing to `core/primitive.array.html`
core/primitive.array.html:341: broken link fragment `#method.to_ascii_lowercase` pointing to `core/primitive.array.html`
core/primitive.array.html:1434: broken link - `core/slice::sort_by_key`
core/primitive.array.html:1519: broken link fragment `#method.sort_by_cached_key` pointing to `core/primitive.array.html`
number of HTML files scanned: 31972
number of HTML redirects found: 9914
number of links checked: 2162137
number of links ignored due to external: 89734

[compiler-errors]

Thanks for the feedback, @eddyb.

For now, I will reopen #92640 which implements a less dramatic set of changes to fix these ICEs? Would you or @lcnr be willing to review that PR instead?

[lcnr]

I misunderstood what exactly is happening here and after @eddyb's comment i am now also not in favor of [T; N] implementing Deref to [T] anymore, though I am not really opposed to it either 😆

will review #92640 and am in favor of only going forward with Deref for arrays in case an unrelated legitimate use case comes up

[nikomatsakis]

I agree with @eddyb that a Deref impl for [T; N] -> [T] is not obviously correct =)

[compiler-errors]

Alright, I'll close this out then