Cache ExerciseDB for offline use and add 4-week refresh prompt

[Appp-maker]

Motivation

• Ensure ExerciseDB content is always available offline by downloading and storing the full dataset locally.
• Avoid live network queries for every search and body-part fetch by switching to a cached local dataset.
• Give users a chance to refresh the local dataset periodically by prompting about refreshes every ~4 weeks.
• Preload the cached dataset on app startup so data is ready when the workout UI is opened.

Description

• Reworked ExerciseDbRepository to download the full exercises dataset, save it to a cache file (exercisedb_cache.json), and serve queries from the cached list using thread-safe Mutex locking.
• Added refresh APIs on the repository: preloadExerciseDatabase(), refreshExerciseDatabase(), and prompt helpers shouldPromptForRefresh() / recordRefreshPromptShown() / recordRefreshPromptDismissed().
• Persisted refresh & prompt timestamps in SettingsRepository via new keys and methods (EXERCISE_DB_LAST_REFRESH, EXERCISE_DB_LAST_PROMPT, saveExerciseDbLastRefresh(), getExerciseDbLastRefresh(), saveExerciseDbLastPrompt(), getExerciseDbLastPrompt()).
• Updated DI to provide OkHttpClient and Gson via new ExerciseDbModule, preloaded the cache from RoutineReminderApp startup, and updated WorkoutViewModel/WorkoutScreen to show a refresh AlertDialog, call refreshExerciseDatabase(), and track prompt state.

Testing

• No automated tests were executed for this change.
• No unit or instrumentation test runs were performed in this rollout.

---

Codex Task