Collect raw conventional-commit data

## Summary
Collect a large corpus of real-world conventional commit messages from open-source repositories.

## Success Criteria
- [ ] Identify target repositories (popular OSS projects using conventional commits)
- [ ] Script to clone/fetch commit history
- [ ] Extract commit messages with metadata (repo, author, date, hash)
- [ ] Store in raw format (before normalization)
- [ ] Target: 100k+ commit messages minimum
- [ ] Document data sources and licensing

## Data Sources (Candidates)
- Angular ecosystem repos
- Vue.js ecosystem
- Commitizen users
- Any repo using semantic-release

## Privacy Considerations
- Only public repositories
- Strip author PII if not needed
- Document data provenance