feat(player): request an App Store review after finishing a movie or episode

[bilipp]

Summary

Lume never prompts users for an App Store rating, even though happy, engaged viewers are the ones most likely to leave a positive review. This adds an automatic, throttled review request that fires at a genuinely positive moment — right after the viewer finishes watching a movie or episode — using SwiftUI's requestReview action.

Implementation

• AppStoreReviewTrigger (Lume/Services/Review/) — a nonisolated enum scalar-flag store (same shape as RecommendationSettings) that owns the throttling policy: it counts finished playbacks and reports eligibility only once ≥ 3 completions accumulate, and at most once per app marketing version (a future update is the soonest it will ask again). StoreKit independently caps the system prompt to three appearances a year, so this is conservative on top of that.
• FullScreenPlayerView — hooks into the existing playback-completion seam in persistProgressDetached, right alongside the one-time Trakt "watched" sync. A Completion is only produced when a movie/episode crosses 90% (never for live channels), so the trigger naturally fires only on meaningful VOD finishes. The requestReview environment action is gated #if !os(tvOS) — tvOS has no App Store rating prompt.

No new user-facing strings (the prompt is system-provided), so no localization changes were needed. The existing manual "Rate Lume" link in Settings is unaffected and complements this automatic path.

Testing

• Acceptance criteria met (throttled prompt fires after meaningful engagement; once per version)
• Tests pass — see note below
• SwiftLint clean (pre-commit hooks passed)
• Tests added — LumeTests/Services/AppStoreReviewTriggerTests.swift covers the threshold, the per-version block, and re-eligibility after an update, using an isolated UserDefaults suite.

Note: The app target builds successfully on the iOS Simulator with this change. The LumeTests target currently has a pre-existing compile error unrelated to this PR — ModelTests/GenreDerivationTests pass a ModelContext where a ModelContainer is expected (these files are unchanged from main). That breaks compilation of the whole test target, so the new AppStoreReviewTriggerTests could not be executed in-suite until that is fixed separately.

Platforms

• iOS / iPadOS
• tvOS (intentionally excluded — no rating prompt API)
• macOS
• visionOS

Related