Build AnnualStatsAggregationService core logic

Structure the service as a class in lib/features/annual_summary/domain/services/annual_stats_aggregation_service.dart. Extract the distribution normalisation into a pure static method _normaliseDistribution(Map counts) → Map and the streak calculation into _calculateStreak(List activityDates, DateTime referenceDate) → int — both can then be tested directly without constructing the full service. For the normalisation remainder fix: compute percentages as doubles, then find the largest bucket and assign it the value needed to make the sum exactly 100.0. For streak: sort dates descending, iterate and compare consecutive pairs using DateUtils.dateOnly() to strip time components before comparison.

Inject dependencies via constructor with default Riverpod providers so integration is seamless; use @riverpod annotation if the project uses riverpod_generator. Consider using Result as the return type instead of throwing — this makes the caller's error handling more explicit, but only adopt this if the rest of the codebase follows this pattern to maintain consistency.

Unit tests (flutter_test) with mocked dependencies. Test groups: (1) Happy path — mock repository returns a known raw map; assert AnnualStatsResult fields match expected computed values including distribution normalisation and streak; (2) Distribution edge cases — single activity type maps to 100.0%; empty activity_type_counts produces empty distribution map; floating-point remainder is handled so values sum to exactly 100.0; (3) Streak calculation — 5 consecutive days = streak of 5; gap on day 3 = streak of 2; no activities = streak of 0; (4) Fallback — repository throws networkError; mock cache returns a prior result; assert getStats returns the cached result; (5) Cache miss — repository throws networkError AND cache returns null; assert getStats rethrows AnnualStatsException. Target 90% line coverage.