Handle RTN projections in assoc type restriction diagnostics

[TaKO8Ki]

Fixes #152887

[rustbot]

r? @jackh726

rustbot has assigned @jackh726.
They will have a look at your PR within the next two weeks and either review your PR or reassign to another reviewer.

Use r? to explicitly pick a reviewer

Why was this reviewer chosen?

The reviewer was selected based on:

• Owners of files modified in this PR: compiler
• compiler expanded to 69 candidates
• Random selection from 10 candidates

[fmease]

View changes since the review

As I said before, could you please further reduce any potential reproducers if they're not minimal enough yet (e.g., by removing unnecessary elements and unnecessary user errors)?

This would be tremendously helpful for other contributors to reason about it. They might want to extend this test file with more test cases years down the line or they see this test failing after modifying the compiler in which case it's very beneficial if the test is easy to digest instead of being a giant blob you first have to wade through. The extra fluff certainly won't increase test coverage by much.

E.g., the code below can be shrunk to the following (maybe even further) while still reproducing the ICE:

#![feature(return_type_notation)]

pub trait Trait { async fn func(); }

impl<T: Trait<func(..): Send>> Trait for T {}

fn check(_: impl Trait) {}

fn main() { check(()); }

[TaKO8Ki]

Oh. Sorry, I completely missed your earlier comment. I’ll reduce the reproducers further.