feat(migrate): sample more pages (incl. institutional) + report organized per page

[JonasJesus42]

From your point that it only captured home/PLP/PDP, and that you'll need the institutional pages — with a single report separated by page.

Changes

• --sample <spec> (default plp=2,pdp=2,other=3,search=1, capped at 15 pages): samples extra pages per kind from the classified sitemap. other = institutional; institutional URLs are preferred (about/contact/policies/…) over deep-category pages that also fall through to other.
• Report organized per page — index.md / index.html / report.html now show a Global components section (captured once) + one section per captured page (home / PLP / PDP / institutional). One report, everything separated by page.
• category-auto no longer lands on an institutional page (institutional keywords filtered from the PLP pick — issue migrate: platform-aware PDP discovery for non-VTEX stores #200).

Verification

Unit tests for parseSample/pickSpread/sampleFromSitemap + the PLP institutional filter. Full suite green (1107), tsc + biome clean, build OK. Live electrolux.com.ec: 3 → 10 pages (home + 3 PLP + 3 PDP + 3 other), report split per page.

Follow-up (noted)

Per-page VTEX block-tree + content (institutional pages' props) — the block tree is currently read on home only; reading it per page is the natural next step to link component ↔ block ↔ content across all pages.

Off main.

🤖 Generated with Claude Code

---

Summary by cubic

Samples more pages by kind (including institutional) and organizes the migrate report by page. Also reads the VTEX IO block tree and CMS content on every captured page and merges them, so content is not limited to home anymore.

• CLI: adds --sample <spec> (default plp=2,pdp=2,other=3,search=1, capped at 15). Samples per kind from the classified sitemap; skips home and already-resolved URLs; for other we prefer institutional URLs (about/contact/policies/…).
• Report: index.md/index.html/report.html now show a Global components section plus one section per captured page, with stable component folder mapping by role+selector.
• VTEX IO: reads the block tree + props on every captured page and merges by treePath; downloads and dedupes content images across pages under a 200-image budget; persists blocks.json and content-assets.json after capture and caches them in capture.json.
• PLP pick: filters institutional keywords so category-auto lands on a real PLP.
• Tests/docs: unit tests for parseSample/pickSpread/sampleFromSitemap and the PLP filter; docs/migrate.md documents per-page VTEX capture and --sample. No breaking changes.

^{Written for commit d1e0e35. Summary will update on new commits.}