Prove 'no visible change' with the repo's visual suite plus a control run

When asked to guarantee a landing-page change is invisible to people, a text diff of the rendered page is weak evidence. Better: run the project's Playwright visual regression suite (toMatchSnapshot against committed baselines) on the changed build, then run the identical suite on the pre-change commit in a git worktree (copy node_modules, use another port) as a control. Failures that appear in both runs are environment or animation-timing noise (typewriter mid-word, moving dots, story card text differ by a few hundred pixels when the page's motion starts at a real-time moment); failures only in the changed build are real. Read the *-diff.png files: Playwright paints only changed pixels in red/yellow over a faded page, so a layout change is obvious at a glance.