Fix repository errors and DeepMind capture

[patrickschiller]

Summary

• verify the configured GitHub repository before HTML or PDF uploads and surface a clear error when it is missing or inaccessible
• add a dedicated Google DeepMind blog DOM adapter that captures the article sections while excluding cover controls and related-post cards
• record DeepMind captures as google-deepmind-dom
• document the behavior in English and German
• add synthetic regression coverage for both bugs

Root cause

GitHub Contents API responses use 404 both for missing archive paths and for an unavailable repository. The existing code treated the first index-path 404 as an empty archive, so repository availability was not established explicitly.

The DeepMind page contains its actual post in top-level main > section elements, while its first article element is a related-content card. The generic fallback selected that card and saved almost none of the post.

User and privacy impact

Users now see an explicit popup error when the configured repository does not exist or the token cannot access it. DeepMind blog posts retain their full article content and omit related-post cards.

No browser permissions or token scopes change. Uploads make one additional read request to the existing GitHub repository metadata endpoint before writing. No storage format or deletion behavior changes.

Checks

• node --test tests/capture-utils.test.js
• python3 -m unittest discover -s tests -p "test_*.py"
• git diff --check
• inspected the live DeepMind page DOM to confirm the article is split across top-level sections and related posts are a separate final section