Link release note issue references and archive pre-v0.230.001 notes

[Paul Lizer (paullizer)]

Why

Release notes referenced issues as bare #1289 or microsoft/simplechat#1145. The notes render on the Jekyll docs site, where kramdown does not auto-link either form, so every reference was dead text for readers.

While mapping references back to issues, a second problem surfaced: the repository default branch is main, but every PR merges into Development. GitHub only honours Fixes #NNNN on merges into the default branch, so no issue has ever auto-closed. 56 shipped issues were still sitting open.

What changed

Docs (this PR)

• Linkified all 165 references inside (Ref: ...) blocks as absolute Markdown links. Every number was classified against the GitHub API first, so the 11 pull-request references point at /pull/ rather than /issues/.
• Backfilled 3 entries whose shipping PR named an issue the note omitted.
• Moved v0.229.063 and older into docs/explanation/archive_release_notes.md — 19 sections, 989 lines, down to v0.190.1. The current page drops from 4,781 to 3,797 lines (452 KB → 391 KB).
• Linked the archive from the bottom of the release notes and from the docs sidebar under Deep reference.

Issues (already applied)

• Closed 40 issues as completed, each with a comment naming the shipping PR, the version, and its release-note entries. Repo open issues went 261 → 221.

Attribution method

Line-level attribution uses each merged PR's own patch of release_notes.md, not the commit → PR API. That API returns every PR whose branch merely contains a commit; open long-lived branch #1228 contains 132 unrelated commits and was initially credited with work spanning 30+ versions it never touched.

Deliberate restraint

• Scoped strictly to (Ref: ...) blocks. The notes also contain prose like **Issue #1 - DOM Manipulation Error** — a list number, not a reference. Blanket linkification would have corrupted it.
• 4 of 7 backfill candidates rejected. Their PRs bundled two unrelated fixes and named both issues, so backfilling would have credited the wrong bullet (e.g. a Data Management scheduler guard tagged with an Analyze-contract issue).
• 296 refs in six pre-convention mega-sections left alone (v0.250.001, v0.241.001, v0.235.003, …). They predate per-PR issue tracking and have no issue to link.
• 11 epics and 5 partial deliveries were not closed — listed below.

Not closed, needs a human call

Issue	Why
#1013 #1014 #1015 #1016 #1017 #1018 #1019 #1020	MCP roadmap tracks/phases; #1013 has 7 sub-issues
#1055 #1058	Mixed-source orchestration epic; #1055 has sub-issues #1056–#1061
#1082	Generalize-workflows epic
#1233	Multi-phase epic (Phases 1–7 plus 7A–7D)
#1047, #358	v0.250.109 explicitly says Closes #1143, related #1047 and #358
#1071	PR #1152 shipped JSON/XML only; universal CSV unconfirmed
#1031	Notes state durable fan-out and automatic continuation are still unavailable

Verification

A validator reconstructs the original file from the two halves, normalizes the links back to their pre-edit spelling, and requires an exact match against HEAD:

lines compared             : 4781
differing lines            : 3 [563, 621, 1091]   <- the reviewed backfills only
content preservation       : OK
issue/pull URL segments    : OK
unlinked refs in Ref blocks: 0
ALL CHECKS PASSED

Docs-only, so no config.py version bump per repo rules.

Follow-up worth filing

The Development vs main mismatch will keep stranding issues. Either change the default branch, or add a workflow that closes referenced issues on merge into Development.