Define normalization rules

## Summary
Define rules for normalizing raw conventional commit data into a consistent format.

## Success Criteria
- [ ] Casing rules defined (lowercase types, scopes, etc.)
- [ ] Whitespace handling documented
- [ ] Character normalization rules (unicode, special chars)
- [ ] Scope canonicalization rules (e.g., aliasing)
- [ ] Rules documented in a format that can be automated

## Context
Raw commit messages vary wildly in formatting. Normalization ensures the n-gram tables and neural model see consistent patterns.

## Open Questions
- Should we preserve original casing for proper nouns in scope?
- How to handle emoji in commit messages?