Avoid a nested `Result<T, PResult<T>>`.

…nkov

coverage: Unexpand spans with `find_ancestor_inside_same_ctxt`

Back in rust-lang#118525 (comment) it was observed that our `unexpand_into_body_span` now looks very similar to `Span::find_ancestor_inside`.

At the time I tried switching over, but doing so resulted in incorrect coverage mappings (or assertion failures), so I left a `FIXME` comment instead.

After some investigation, I identified the two problems with my original approach:
- I should have been using `find_ancestor_inside_same_ctxt` instead, since we want a span that's inside the body and has the same context as the body.
- For async functions, we were actually using the post-expansion body span, which is why we needed to forcibly set the unexpanded span's context to match the body span. For body spans produced by macro-expansion, we already have special-case code to detect this and use the pre-expansion call site as the body span. By making this code also detect async desugaring, I was able to end up with a body span that works properly with `find_ancestor_inside_same_ctxt`, avoiding the need to forcibly change the span context.

…, r=WaffleLapkin

refactor(liveness): move walk_expr outside of every match branch

…-errors

Simplify Parser::ident_or_error

Avoid a nested `Result<T, PResult<T>>`.