Refactor samples into an accessible static learning gallery

[zhaowl94]

Summary

• preserve every existing HTML entry point while adding a bilingual root gallery
• rebuild the profile, Ultimate, and conference exercises with semantic, responsive, framework-free HTML/CSS
• keep direct file://, local HTTP, and GitHub Pages project-path usage compatible
• add local-only conference registration behavior with no requests, cookies, or browser storage
• add Node 24 tooling, pinned GitHub Actions, Pages deployment, Dependabot, documentation, licensing, and generated local imagery
• add all verification under test/, including Windows Edge accessibility-tree, forced-colors, 200%-width, direct-file, visual, Axe, keyboard, form, security, unit, and static checks

Windows compatibility

The repository does not modify Windows environment variables, PowerShell execution policy, or global Git/npm configuration. Windows instructions use npm.cmd so restricted PowerShell environments remain supported.

Verification

• npm.cmd run check: 10 HTML and 7 CSS files passed
• npm.cmd run test:unit: 7/7 passed
• npm.cmd run test:e2e:windows: 163/163 passed across installed Chrome, Edge, Firefox, and mobile Chromium
• isolated Ubuntu npm ci: 359 packages, 0 vulnerabilities
• isolated Ubuntu npm run test:e2e:linux: 142/142 passed across Chromium, Firefox, WebKit, mobile Chromium, and five visual baselines
• direct file:// regression: 10/10 passed in Chromium and Edge
• Lighthouse: all four categories scored 1.00 on five representative pages
• five external links returned HTTP 200
• workflow YAML parsed successfully and all 16 action references match official 40-character tag SHAs

Accessibility boundary

Repeatable Windows Edge accessibility-tree, forced-colors, keyboard, and 200%-equivalent-width checks passed. Direct Narrator speech output could not be automated safely and remains explicitly documented as a manual verification limitation in test/manual-accessibility.md.