discovery: walk Squarespace blog archive via ?format=json-pretty

[davipontesblog]

What this changes

Every Squarespace blog section exposes itself as a paginated JSON feed at
<blog-prefix>?format=json-pretty[&offset=<addedOn>]. Each page returns ~20
items with title, urlId, fullUrl, publishOn, addedOn, categories, tags,
assetUrl (cover image), authorId, and full author name. The endpoint
works without auth, on both 7.0 and 7.1, and crucially still serves real
metadata on 7.1 Fluid Engine sites where per-URL ?format=json returns an
empty body.

Adds to squarespaceAdapter:

• SqsBlogArchiveEntry type
• detectBlogPrefixes() — finds blog sections by date-based URL pattern
  (/foo/YYYY/M/D/slug) or known conventions (/blog, /journal, /news,
  /posts, /stories)
• fetchBlogArchive() — paginates via ?offset=, dedupes by urlId, stops
  on empty page / non-advancing offset / 200-page safety cap
• discover() merges any post URLs the sitemap missed and stashes per-post
  metadata on SquarespaceInventory.blogArchive
• extract() builds an archiveByUrl map inside extractPage() and uses
  the archive entries as a fallback for assetUrl, categories, tags,
  publishOn, and author display name when per-URL JSON is empty

How I found it

While migrating walkaboutchronicles.com (1,500+ post 7.1 blog, two
contributors). The per-URL probe path would have needed ~3,000 requests for
author reattribution + featured-image rebuild; the archive walker did it in
~75. Author IDs from the archive disambiguated two contributors who shared a
display name format.

Tested against

• Real site (analogous PHP port, then verified against the live
  walkaboutchronicles.com archive feed)
• Tests pass (npx vitest run — 409 passed, 2 skipped)
• npx tsc --noEmit clean
• 5 new unit tests for detectBlogPrefixes (date-based, conventional,
  ordering, threshold, no-match)
• No new dependencies

Discovery log entry added to DISCOVERIES.md

• Yes