Implement Badge Evaluation Service

Structure the service as a Dart class BadgeEvaluationService with a single public async method evaluate(). Fetch all required data in parallel using Future.wait([getActiveDefinitions(), getEarnedBadgeIds(mentorId), getStats(mentorId)]) before entering the evaluation loop — never fetch inside the loop. Implement criteria evaluation as a strategy/visitor pattern: a BadgeCriteriaEvaluator interface with concrete implementations per criteria type (SessionCountCriteriaEvaluator, StreakCriteriaEvaluator, etc.) — this makes adding new criteria types trivial and keeps each evaluator independently testable. Convert earned badge IDs to a Set before the loop for O(1) exclusion lookups.

If Riverpod is used in the dependency injection layer, register BadgeEvaluationService as a Provider or FutureProvider for easy override in tests. Never log badge criteria thresholds or stats values at info level; use debug-only logging.

Write unit tests (flutter_test): (1) a mentor with qualifying stats for badge A and not badge B returns only badge A's ID, (2) already-earned badges are excluded even when stats still qualify, (3) empty stats returns empty list, (4) zero active definitions returns empty list, (5) each supported criteria type (session count, streak, activity type) is evaluated correctly for both pass and fail cases. Use mockito to mock badge-definition-repository, badge-repository, and peer-mentor-stats-aggregator. Achieve ≥ 90% branch coverage measured with flutter test --coverage. Write at least one integration test connecting to a local Supabase test instance to verify end-to-end evaluation flow.