Fix sitemap lastmod regression from PR #539 merge

[timujinne]

Summary

PR #539 (V117 migration) was merged with a silent sitemap regression. The 3-way merge overwrote two sitemap helpers that existed on upstream/dev but had been removed long ago on the source fork — restoring them.

What broke after PR #539

File	Before merge	After merge
lib/modules/sitemap/sources/publishing.ex	lastmod: latest_post_date(slug, language)	lastmod: nil
lib/modules/sitemap/sources/static.ex (lines 202, 227)	lastmod: static_lastmod(path)	lastmod: Date.utc_today()
defp latest_post_date/2	present	deleted
defp static_lastmod/1	present	deleted

SEO impact of the regression:

• Publishing group listing pages (/blog, /posts, …) lost their <lastmod> element entirely
• Static homepage / now reports lastmod: today's date on every sitemap regeneration — a known SEO anti-pattern

Fix

Restores both files to their state at 0b561d1b — the last upstream/dev commit before the PR #539 merge. Both helper functions (latest_post_date/2, static_lastmod/1) and their callsites come back.

Original commits that introduced this logic:

• e7b0ef60 Add lastmod to sitemap group listings and homepage
• 54b9b899 Add lastmod to sitemap router-discovered and static entries

Test Plan

• mix compile --warnings-as-errors clean
• mix docs clean
• mix quality (credo + dialyzer) clean
• Pre-commit hook passed
• Verify sitemap XML output on a parent app — group listing pages should emit <lastmod> again, homepage <lastmod> should reflect latest publishing content rather than today's date

Context

Detailed pre-merge analysis: dev_docs/pull_requests/2026/539-v117-document-composition-tables/CLAUDE_REVIEW.md (already on dev via commit 17b3b6e1) — see section "#1 Undeclared scope: PR also reverted the dynamic-lastmod sitemap logic on dev".

[ddon]

Thanks for catching this @timujinne — your diagnosis matches exactly what the post-merge review of #539 turned up, and the citation to CLAUDE_REVIEW.md (section #1 "Undeclared scope") is the right pointer.

Closing this in favor of 29fe377c, which already landed on dev and does the same restore plus a small perf improvement:

• Sources.Publishing.latest_post_date/2 — restored verbatim (drives per-group-listing <lastmod>), same as your PR.
• Sources.Static.static_lastmod/1 — restored, same call-sites (static.ex:202,227), but "/" routes through a new Sources.Publishing.latest_post_date_global/0 instead of Publishing.collect/1.
• Sources.Publishing.latest_post_date_global/0 — new public helper. Old shape called Publishing.collect/1 from static_lastmod("/") just to discard everything except :lastmod, which itself triggers ~3× list_posts/2 calls per group inside collect/1 (via group_has_posts_for_language?, latest_post_date, and collect_group_posts). New shape does one list_posts/2 walk per group, returns max published_at. Semantically equivalent to max(collect/1 entries' lastmod) since group-listing :lastmod IS the per-group max post date.
• CHANGELOG entry under 1.7.110### Fixed.

mix precommit clean. No test coverage on either helper before or after — same gap as pre-merge dev. If you want a follow-up PR to add LiveView-test-style coverage on the sitemap sources I'd be glad to review.

Sorry for the duplicate work — both branches were investigating in parallel and ours just landed first.