docs: groom the unreleased changelog ahead of the next release

[kkozik-amplify]

Why

Two outside contributors sent the five fixes currently sitting in Unreleased, and the changelog didn't say so. This repo has credited contributors by name since at least 6.x — dropping it now would be a poor look on a release that is largely their work.

Opened as a draft as a holding branch for changelog grooming before the next release, so anything else we want to tidy can land here rather than as a scatter of one-line PRs.

What

Contributor credit

Follows the existing convention exactly — Thanks, @user before a single trailing link (see the 7.2.1 and 7.2.0 sections):

Entry	Credit	PR
Empty heredoc	@livingstaccato	#312
Negative integer literals	@livingstaccato	#311
strip_string_quotes — expression interiors	@livingstaccato	#313
strip_string_quotes — escape sequences	@livingstaccato	#313
CRLF line endings	@agu2347	#317
Heredoc trailing whitespace	@agu2347	#318

@livingstaccato also reported #307–#310, so the credit covers both halves of that work.

Link style

The Unreleased entries pointed at issue numbers; every other entry in the file points at the PR. The file is 67 /pull/ links to 22 /issues/, and all 20 existing Thanks lines use /pull/. Switched the seven Unreleased entries to match, including the py.typed line (#298 → its PR #299).

Note the two strip_string_quotes entries now both point at #313. That's correct — one PR fixed two reported issues, and they're worth listing as separate behaviour changes.

Deliberately not done

• No credit line for issue reporters who didn't send a PR. @rouge8 reported py.typed deleted from package #298 and the fix was internal (restore py.typed marker (#298) #299). Every existing Thanks in this file goes to a PR author, and inventing a Reported by convention felt like the wrong thing to slip into a grooming PR. Easy to add if we want it — say the word.

Unrelated thing worth a look

All six issues fixed by these PRs — #307, #308, #309, #310, #315, #316 — are still open, even though each PR carried a recognised Fixes #N (the API reports them under closingIssuesReferences) and all merged into main over 12 hours ago. Auto-close appears not to have fired. Not something this PR touches, but they probably want closing by hand.

Test plan

• pre-commit run — clean, including mdformat
• Verified every link resolves: restore py.typed marker (#298) #299, fix: load negative integer literals as numbers (#307) #311, fix: parse heredocs with an empty body (#309) #312, fix: make strip_string_quotes yield values, not broken source (#308, #310) #313, Fix CRLF (\r\n) line endings failing to parse #317, Fix heredoc bodies losing trailing blank lines/spaces to a greedy rstrip #318 all exist and are MERGED
• Verified every attribution against the actual PR author via the API
• Docs only; no code touched