trait_sel: sizedness goals prefer alias candidates

[davidtwco]

Fixes #143992

• abd07de: Reverts trait_sel: MetaSized always holds temporarily #144016 so that MetaSized bounds are checked properly, and updates all the tests accordingly, including making tests/ui/sized-hierarchy/incomplete-inference-issue-143992.rs fail when it shouldn't
• 90e61db: Prefer alias candidates over parameter environment candidates for sizedness, auto and default traits. tests/ui/sized-hierarchy/incomplete-inference-issue-143992.rs passes again, but tests/ui/generic-associated-types/issue-93262.rs starts failing when it shouldn't
• e412062: No longer require that predicates of aliases hold in well-formedness checking of the alias. tests/ui/generic-associated-types/issue-93262.rs passes again

Each commit updates all the tests to their new output so it should be easy enough to see what the impact of each change individually is. After all of the changes, tests that pass when they didn't before or vice versa:

• tests/ui/extern/extern-types-size_of_val.rs
  • Previously passing, but only because of trait_sel: MetaSized always holds temporarily #144016, now correctly errors
• tests/ui/sized-hierarchy/incomplete-inference-issue-143992.rs
  • Previously failing on next solver, only because trait_sel: MetaSized always holds temporarily #144016 only applied to the old solver, passing now with 90e61db
• tests/ui/sized-hierarchy/overflow.rs
  • Previously passing, but only because of trait_sel: MetaSized always holds temporarily #144016, now correctly errors
• tests/ui/generic-associated-types/issue-92096.rs
  • Previously passing, due to e412062
  • Fails to prove C::Connecting<'placeholder>: Send which is required when proving that the generator is Send. This is an instance of Tracking issue for incorrect lifetime bound errors in async #110338.
• tests/ui/higher-ranked/trait-bounds/normalize-under-binder/norm-before-method-resolution-opaque-type.rs
  • Previously passing, now failing in the next solver, due to 03e0fda
  • Expected that this test now fails as ambigious, see Zulip

This had a crater run in #142712 (comment) alongside some other changes.

r? @lcnr
cc #142712 (this extracts part of that change)

[lcnr]

wrt 6aa3732 it's not that we stop to normalize the predicates of projections, it's that when comfirming alias-bound/Projection candidates, we required that the alias is well-formed. While we need to normalize these obligations before proving them, the normalize isn't the relevant part, the "we need to prove them at all" is.

wrt tests/ui/higher-ranked/trait-bounds/normalize-under-binder/norm-before-method-resolution-opaque-type.rs see https://rust-lang.zulipchat.com/#narrow/channel/144729-t-types/topic/sizedness.20bounds.20in.20explicit_implied_predicates_of.20.28.23142712.29/near/526987384. It's intended that this test changes to be ambiguous.

tests/ui/generic-associated-types/issue-92096.rs fails to prove C::Connecting<'placeholder>: Send which is required when proving that the generator is Send. This is 'just' an instance of #110338

Did you already crater run this behavior?

[lcnr]

I dislike this as is 🤔 maybe have a shared used by both old and new solver (with comments)

enum CandidatePreferenceMode {
    Default,
    Marker,
}

impl CandidatePreferenceMode {
    fn compute<I: Interner>(cx: I, trait_def_id: I::DefId) -> CandidatePreferenceMode;
}

[davidtwco]

Fixed this, introduced a CandidatePreferenceMode and used it in both solvers

[davidtwco]

Did you already crater run this behavior?

We did a crater run in #142712 (comment), which had all of these changes and included implying MetaSized on associated types.

[davidtwco]

Updated the README and corrected the descriptions of everything per your comment