Summary

• Adds resolveBlockingGroupKeys, combining the independent URL-path (derivePathGroupKey) and stylesheet (deriveStylesheetGroupKey) blocking keys added in feat(page-cluster): add URL-path and stylesheet blocking-key derivation #902 into one final grouping key per page.
• Priority-with-fallback: prefer the stylesheet signal when it's shared by at least minCssGroupSize pages (default 2), otherwise fall back to the URL-path key.
• Reuses computeDocumentFrequency/splitTokensByFrequency (feat(page-cluster): add frequency-based template/content token split #901) to strip stylesheets common across most of the batch (e.g. a shared reset/font file) before hashing, so two unrelated pages that only share such a file aren't wrongly merged into one group.
• Design was checked against entity-resolution blocking literature (DNF blocking, canopy clustering, ensemble blocking) before implementation; see JSDoc for the reasoning (in particular why OR-merging the two signals would be mathematically unable to reproduce the desired splitting behavior).
• Validated against two real crawl archives (a homogeneous single-template site and a large multi-section site) in addition to unit tests.

Test plan

• yarn build
• yarn lint
• yarn test (1180 tests passing)
• /code-review xhigh (6 findings, all fixed)
• /qa-engineer review
• /product-manager review
• /doc review