feat: improve content extraction and reading experience

[technowizard]

Summary

Improves content reliability from extraction through reading, while smoothing RSS, theme, and development-server behavior.

What changed

• Wait for detected AWS WAF challenges before extracting article content.
• Harden HTML-to-Markdown conversion to:
  • preserve valid article content and multilingual text
  • resolve relative and fragment links against the publisher
  • discard active-content link schemes
  • remove only duplicate title headings
  • produce valid linked-heading Markdown
• Restrict reader data URLs to raster image sources.
• Treat zero RSS subscriptions as a normal first-use state without requesting feed items.
• Align RSS heading typography with Home.
• Persist User Menu theme selections through the existing settings-backed preference flow.
• Disable the global rate limiter in development while retaining endpoint-specific limits for sensitive operations.
• Add regression coverage at each changed boundary.

Verification

• pnpm --filter web test --run — 144 tests passed
• pnpm --filter server test --run — 315 tests passed
• pnpm --filter web typecheck
• pnpm --filter server typecheck
• pnpm --filter web build
• pnpm --filter server build
• pnpm lint
• git diff --check
• pnpm --filter server test --run src/lib/create-app.test.ts — passed during implementation
• Commit hooks passed Oxlint and Oxfmt for the final rate-limit change