Playground redesign and SEO metadata

[includeamin]

Redesigns the web playground as a three-pane workbench and adds SEO metadata so the playground can show up in search results and render rich link previews.

Design

• Layout: a full-height layout with three panes:
  • a file sidebar;
  • an editor with tabs for openapi-aggregator.yaml and the selected file;
  • output tabs: Merged spec, API reference, and Problems with a count badge.
• Top bar: a logo mark, a tagline, a styled example picker and the "Copy share link" action. Below it, a dismissible intro strip links to installing the CLI and to how merging works; dismissing it is remembered per browser.
• Status bar: shows the run state (Merged, Merged with warnings, Config has errors, Merge failed, Waiting for permission to fetch), each source, the problem count and the output format.
• Source colours: each source gets a colour. It appears as a dot next to its file in the sidebar (hollow if the config doesn't use the file) and as a chip in the status bar.
• Fonts and theming:
  • IBM Plex Sans and Mono are bundled with @fontsource, so no request goes to a font CDN.
  • The editors draw their colours from the page's CSS tokens, using One Dark syntax colours in dark mode.
• Scalar API reference:
  • It follows the system light or dark setting and uses the page fonts.
  • Its developer toolbar, "Ask AI" agent and telemetry are turned off.
• Small screens: below 900 px the panes stack vertically and the file actions are always visible.

The run-state and colour logic lives in the new web/src/sources.ts, with unit tests.

SEO

• Search snippet: a title and meta description aimed at searches like "merge OpenAPI specs" and "Swagger merge", plus a canonical URL and keywords.
• Link previews: Open Graph and Twitter card tags with a new 1200×630 preview image, web/public/og-image.png.
• Structured data: SoftwareApplication JSON-LD describing a free developer application, with its repository and license.
• No-JS fallback: a visually hidden <h1> and a <noscript> summary for crawlers that don't run JavaScript.
• Sitemap: sitemap.xml, to submit in Google Search Console after deploy. There is no robots.txt: for a project site, crawlers only read it at the domain root.

Separately, the repo's description, homepage URL and topics have been updated on GitHub.

Testing

• 55 vitest unit tests pass (1 skipped because it needs the network), and typecheck is clean.
• All 4 Playwright end-to-end tests pass. The problem badge now shows 1 instead of (1).
• Checked with screenshots in light mode, dark mode and at 390 px wide. There is no horizontal scroll and no console errors.