feat: /simplify should cross-reference prior codebase review findings

Problem

The /simplify skill reviews the diff for code reuse, quality, and efficiency issues within the changed code. However, it does not cross-reference against findings from prior codebase reviews (e.g., from codebase-review-agent) to check whether new code repeats known issues.

Example

During a session working on the wger-flutter project:

1. The initial codebase-review-agent run identified that the app uses localized strings (AppLocalizations) throughout the codebase
2. New code was written with hardcoded English strings (e.g., 'Health sync', 'Import weight from Apple Health or Health Connect')
3. /simplify was run and flagged several valid issues (duplicate date logic, O(n*m) dedup, redundant comments) but did not flag the hardcoded strings
4. The user manually caught the l10n issue

The /simplify agents correctly reviewed the diff in isolation, but missed that the new code introduced the same class of problem the codebase review had already identified.

Suggestion

The /codebase-review-agent should save the assessment as a review document.

When a codebase review document exists (e.g., in wingspan/reviews/), the /simplify skill (or one of its sub-agents) could:

1. Read the most recent codebase review findings
2. Check whether the diff introduces code that violates any of the documented conventions or repeats known issues
3. Flag these as "previously identified convention" violations

This would close the gap between "what we know about the codebase" and "what we check in new code."

Start with the /simplify skill and /codebase-review-agent entry points, then inspect how review documents in wingspan/reviews/ are created and discovered. Define how the most recent findings are selected and how /simplify reports violations as previously identified conventions. Done means a review document can inform a subsequent /simplify run without breaking its existing diff checks.