Reuse querySelectorAll for getElementsByTagName

[andrewiggins]

Summary

Builds on #719. This layer contains only the tag-name lookup refactor and its required selector support.

• Replace the dedicated getElementsByTagName() descendant loop with a structured matcher passed to querySelectorAll().
• Introduce an internal qualified-name matcher: DOM tag-name lookup compares the full qualified name, while CSS type selectors continue comparing localName.
• Supply precomputed qualified names through htmlName rather than value, extending the normalized-name matcher contract and type tests.
• Preserve wildcard lookup, string conversion, ASCII-only HTML query normalization, and case-sensitive foreign-name matching.
• Return the existing static NodeList, including item() access; this does not implement a live HTMLCollection.

This shares traversal and matching code; it is not an overall performance-speedup claim. Earlier synthetic Node/V8 measurements of this approach remained slower than the dedicated loop. Those measurements were not a final-commit or browser benchmark. Plain-array results and scalar matcher experiments are not included.

Validation

• Full polyfill unit suite: 434 tests passed across 30 files.
• Polyfill TypeScript check passed.
• Prettier and diff checks passed.
• Added structured-matcher coverage verifying that HTML matching uses htmlName, foreign matching uses the original name, and neither uses value.

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}