Add offline cache for scenario rules

Check the existing codebase for any established caching utilities or storage abstractions before introducing a new pattern. If Hive is already a dependency, use a dedicated Hive box named 'scenario_rules_cache'. If only flutter_secure_storage is available, serialize the List to JSON string and store under key 'scenario_rules_{chapterId}'. Implement a CachedRuleSet wrapper model: `{ rules: List, cachedAt: DateTime, chapterId: String }`.

For the cache-first + background refresh pattern, return a Stream> from fetchRulesForChapter — yield cached value immediately, then yield refreshed value when background fetch completes. Alternatively, expose an explicit refreshRules() method and rely on Riverpod state for reactivity. The chapter change invalidation can be triggered by listening to an authentication state change provider. Ensure the 24-hour TTL is calculated server-time-agnostic using DateTime.now() on device.

Unit tests with flutter_test: (1) when cache is valid, fetchRulesForChapter returns cached rules without calling Supabase; (2) when cache is empty, Supabase is called and result is cached; (3) when cache is expired (timestamp > 24h), Supabase is called synchronously and cache is refreshed; (4) when chapter changes, old cache is invalidated; (5) background refresh failure does not throw — stale cache is retained; (6) cache key is chapter-scoped (rule set for chapter A not returned for chapter B request); (7) logout clears all cached rules. Use a fake/mock for the underlying storage (Hive or flutter_secure_storage) and a mock for the network repository. Integration test: install app, fetch rules while online, enable airplane mode, restart app, assert rules still available.