fix: improve GDPR data quality and simplify article structure

[KevinRabun]

Summary

This PR fixes data quality issues in the bundled GDPR text data and simplifies the article structure for improved usability.

Changes

Data Quality Fixes

• Remove navigation artifacts from all 99 articles and 173 recitals (patterns like ← Art. N GDPR, Table of contents, Report error, ← Recital N Recital M → All recitals)
• Fix recital 1 edge case - first recital has different navigation pattern without leading ←
• Strip standalone paragraph numbers (1, 2, etc.) from article text for clean prose

Structure Simplification

• Remove paragraphs array from articles - the source website (gdpr-info.eu) doesn't cleanly separate main paragraph numbers from sub-paragraph numbers, making reliable extraction impossible
• Simplify article structure to just: number, title, text
• Reduce JSON size from ~500KB to ~370KB (26% reduction)

Before

{
  "number": 33,
  "title": "Notification of a personal data breach...",
  "text": "1\nIn the case of...\n2\nWhere the notification...",
  "paragraphs": [{"number": 1, "text": "..."}, {"number": 2, "text": "..."}]
}

After

{
  "number": 33,
  "title": "Notification of a personal data breach...", 
  "text": "In the case of a personal data breach, the controller shall..."
}

Testing

• All 294 tests pass
• Verified all articles and recitals are artifact-free
• Verified definitions extraction still works from Article 4