docs(skills): preserve glyphs in text reveal masks

[user-github-me]

Tight line-height with an overflow: hidden text mask can cut descenders and accents after a reveal has finished. Document mask padding with a compensating negative margin, explain why the mask must use the text's font size for em units, and link the guidance from the mask-reveal-up adapter.

Also describe releasing clipping on the timeline after the entrance and explain that data-layout-allow-overflow suppresses resting-state checks. Regenerate the skills manifest.

Refs #5231. This addresses the requested skills guidance; the reported audit behavior remains for a separate change.

Validation:

• Chrome 152, local Inter 600: compared masked and unclipped screenshots for Betreuung, Ägypj, and Épýj at two settled timestamps. The unpadded mask lost 1,048–3,211 visible pixels; padding preserved the complete image byte-for-byte and kept the original text position. The initial parked text remained hidden.
• Both local compositions passed hyperframes lint and hyperframes check --strict --at 1.2,1.8, confirming the current audit blind spot.
• Repository lint, final skills cross-reference checks, CLI build, and formatting passed.

[jrusso1020]

Useful guidance. A tight display line-height with an overflow: hidden mask does clip descenders and accents at rest, and the allow-overflow attribute waives the resting state too, so check alone won't catch it. Padding with a compensating margin is a sound approach, and the reminder to keep the parked line outside the enlarged mask is the right caveat. Another open skills docs change also updates the hyperframes-animation manifest hash, so whichever lands second will need a regenerate.

— Rames