[fmease]

1. Don't artificially reject duplicate (non-empty) where clauses on bodiless type aliases / associated types.
   • We used to do so due to overzealous parse error recovery code that was added in PR better suggestion for duplicated where clause #93571 (submitted & merged: Feb 2022) at basically the same time as trailing where clauses for type aliases were introduced in PR Change location of where clause on GATs #90076 (submitted: Oct 2021, merged: Mar 2022)
   • Note that the Rust Reference currently incorrectly states WhereClause? ( = Type WhereClause? )?
     • it needs to be updated to say WhereClause? ( = Type )? WhereClause?
     • I'll do that separately
   • Found thanks to [@]matthiaskrgr fuzzing rustc against rasur and reporting the incongruence downstream: cfg(false) where where fmease/rasur#24.
2. Reject (literally) where where (i.e., duplicate empty where clauses) on const items (unstable feature generic_const_items). See commit for details.
3. Tiny compiler & test cleanups that were necessary to soundly navigate these seas

Best reviewed commit by commit.

_{(No LLM was or will be used by me during the entire creation process of this PR)}

[rustbot]

r? @beetrees

rustbot has assigned @beetrees.
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 77 candidates
• Random selection from 22 candidates

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.