Code Review Without Borders: Evaluating Synthetic vs. Real Data for Review Recommendation