Remove incorrect parse error recovery code that mistakes as casts for the long removed type ascription

[fmease]

Back when we still had type ascription syntax $expr : $ty, parse_assoc_op_cast would parse both as casts & type ascription.

During that time (namely in commit 8c5dafd), parse error recovery from code like label: loop {} was added (label lacks leading apostrophe). However, it never checked if we did actually parse a : and not an as meaning to this day we emit a nonsensical diagnostic for expressions like label as loop {}! This PR does away with this code & further cleans up in the area (thanks to type ascription being gone).

In case you're wondering, we do still recover from expr stmts like label: loop {} as we have some code in the stmt parser for this.

Since the removal of the type ascription syntax we do indeed no longer provide that recovery for arbitrary exprs (e.g, (label: loop {})) which I find absolutely acceptable.

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

[rustbot]

r? @fee1-dead

rustbot has assigned @fee1-dead.
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, parser
• compiler, parser expanded to 76 candidates
• Random selection from 21 candidates

[chenyukang]

link #101728

[chenyukang]

maybe add a ui test for it since there is a observable change on diagnostics.

[fmease]

maybe add a ui test for it since there is a observable change on diagnostics.

I'm usually pro tests (of course) but in cases like here (or over there: #162704) I'm less in favor since the larger context is me being on a spree to dejank the parser trying to shave off as much crusty code as possible (cc #162269, #161796).

How likely would it be for label as loop {} to regress again? I believe it's 0% because we've de-RFC'ed $expr : $ty as you obviously know. So what would the odds be? It's just a lot more satisfying to have diffs where removed>>added ^^' Adding a regression test that in my eyes doesn't bear any value would make a dent in that.