feat(map): return discovered sitemap document URLs

[us]

/v1/map walked the sitemap tree during discovery but returned only the
<url><loc> page entries. The sitemap documents themselves were dropped on the
way out, so a caller had no way to get /sitemap.xml, /sitemap_index.xml or
/product-sitemap.xml back. They are already in hand at that point, so
surfacing them costs no extra fetch.

What changed

• fetch_sitemap_tree returns SitemapWalk { pages, sitemaps }. Only documents
  that answered with parseable content are listed, so a fallback path that 404s
  (/wp-sitemap.xml on a non-WordPress site) is never reported as a sitemap of
  the site.
• MapData gains sitemaps: string[]. Kept out of links because a sitemap
  file is not a page, and merging them would silently change the output of every
  existing caller. Empty when useSitemap is false.
• CLI: text mode prints Sitemap: <url> lines to stderr so stdout stays a
  pipeable page list; --format json gains the field.
• Docs: map.md, response-shapes.md, llms-full.txt, both OpenAPI specs
  (the embedded crates/crw-server/openapi/ copies and the docs/ copies), and
  the regenerated static pages.
• docs/features/map/index.html documented /v1/map with the /v2 envelope
  (top-level links) and parameters /v1 does not accept (search,
  includeSubdomains, ignoreSitemap). Its Python example would have raised
  KeyError against the real response. Corrected in the same pass.

Compatibility

The Firecrawl-compat /v2/map shape is untouched: it builds its own
V2MapResponse / V2Link types and does not read MapData. Verified against a
running server: top-level keys stay ["links", "success"] and each link object
stays {"url": ...}.

sitemaps is additive on /v1, so existing /v1 callers are unaffected.

Verification

Against a local server on the built binary:

• POST /v1/map {"url": "https://wordpress.org", "crawlFallback": false, "limit": 5}
  returns 27 sitemaps (sitemap.xml, news-sitemap.xml, themes/sitemap.xml,
  plugins/sitemap-index-1.xml, ...) alongside the page links.
• POST /v1/map {"useSitemap": false} returns "sitemaps": [].
• POST /v2/map shape unchanged.
• A site with no sitemap (python.org: /sitemap.xml 404, no Sitemap: line in
  robots.txt) returns an empty list rather than a fabricated one.

New unit test covers the index-plus-child case and asserts a 404 seed is not
reported. OpenAPI drift guard run against a release build:
live /openapi.json matches docs/openapi.json.

ref #440