fix(core/lint): findRootTag ignores <video>-like tokens inside HTML comments / <style> / <script>

[ukimsanov]

Summary

A literal <video> token written inside a CSS comment block (e.g. /* card uses <video> as surface */) was being picked as the composition root, producing two cascading false errors: root_missing_composition_id + root_missing_dimensions.

findRootTag now masks <!-- --> / <style>…</style> / <script>…</script> ranges with same-length whitespace (so tag offsets stay aligned for downstream consumers) before running the tag scan.

Verification

• Existing stripJsComments / extractScriptTextsAndSrcs exports preserved (downstream rule consumers in rules/{adapters,composition,core,gsap}.ts read rootTag.raw + rootTag.index against the real source offsets — safe under the new masking)
• bun run --filter @hyperframes/core test passes

Sibling PRs in this stack

• fix(producer): shim CJS __dirname / __filename in ESM banner #1445 — fix(producer): __dirname ESM banner shim
• feat(cli): capture-video on-demand fetcher + capture pipeline robustness #1447 — feat(cli): capture-video on-demand fetcher + capture pipeline robustness

All three independent; merge order doesn't matter.

Test plan

• bun run --filter @hyperframes/core typecheck passes
• bun run --filter @hyperframes/core test passes

[ukimsanov]

Closing — verified no composition in current main has the trigger pattern (<video> token literally written inside a CSS comment / <style> / <script> block). Synthetic repro works but no real user hits it. Re-open if someone reports it.