v0.3.0 — Phase β-2 Study-Loop

Voller Lern-Flow mit Web-Parität: fällige Karten via /reviews/due
laden, flip + rate (4 Buttons + Haptic), Grades via Offline-Queue
ans Server-FSRS schicken.

- Card/Review/DueReview DTOs mit snake_case + camelCase-deckId-
  Sonderfall im embedded card-Subobjekt
- CardType-Enum (alle 7 Typen), Rating-Enum mit deutschen Labels
- Cloze-Helper 1:1-Port aus cards-domain (extractClusterIds,
  subIndexCount, clusterId, renderPrompt/Answer, hint)
- CardsAPI.dueReviews(deckId:) + gradeReview(cardId,subIndex,rating,reviewedAt)
- PendingGrade SwiftData-Model + GradeQueue (FIFO-Drain, originaler
  Timestamp bleibt, bei Netzfehler in Queue, Retry beim nächsten Drain)
- StudySession @Observable State-Machine
- CardRenderer für basic, basic-reverse, cloze; Placeholder für
  image-occlusion/audio-front/typing/multiple-choice (β-3/β-4)
- RatingBar mit UIImpactFeedbackGenerator (medium/heavy)
- StudySessionView per NavigationLink aus DeckListView
- 9 neue Tests (Cloze: 8, Review-Decoding: 3), insgesamt 17 grün

Server-authoritative FSRS bleibt — kein ts-fsrs-Port.
Endurance-Test auf realem Gerät steht aus (siehe PLAN.md).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Tests/UnitTests/ReviewDecodingTests.swift

@@ -0,0 +1,83 @@
+import Foundation
+import Testing
+@testable import CardsNative
+
+@Suite("Review-JSON-Decoding")
+struct ReviewDecodingTests {
+    @Test("Review-Wire-Format decodet vollständig")
+    func decodesReview() throws {
+        let json = """
+        {
+            "card_id": "card_1",
+            "sub_index": 0,
+            "user_id": "user_1",
+            "due": "2026-05-13T10:00:00.000Z",
+            "stability": 2.5,
+            "difficulty": 5.0,
+            "elapsed_days": 1.0,
+            "scheduled_days": 3.0,
+            "learning_steps": 0,
+            "reps": 5,
+            "lapses": 1,
+            "state": "review",
+            "last_review": "2026-05-10T10:00:00.000Z"
+        }
+        """.data(using: .utf8)!
+
+        let decoder = JSONDecoder()
+        decoder.dateDecodingStrategy = .iso8601withFractional
+        let review = try decoder.decode(Review.self, from: json)
+
+        #expect(review.cardId == "card_1")
+        #expect(review.subIndex == 0)
+        #expect(review.state == .review)
+        #expect(review.reps == 5)
+        #expect(review.lastReview != nil)
+    }
+
+    @Test("DueReview embedded card decodet (camelCase deckId!)")
+    func decodesDueReview() throws {
+        // Achtung: Server liefert hier `deckId` camelCase im embedded card,
+        // weil das aus Drizzle direkt rauskommt, nicht durch toCardDto.
+        let json = """
+        {
+            "card_id": "c1",
+            "sub_index": 0,
+            "user_id": "u1",
+            "due": "2026-05-13T10:00:00.000Z",
+            "stability": 0,
+            "difficulty": 0,
+            "elapsed_days": 0,
+            "scheduled_days": 0,
+            "learning_steps": 0,
+            "reps": 0,
+            "lapses": 0,
+            "state": "new",
+            "last_review": null,
+            "card": {
+                "id": "c1",
+                "deckId": "d1",
+                "type": "basic",
+                "fields": {"front": "Was ist 1+1?", "back": "2"}
+            }
+        }
+        """.data(using: .utf8)!
+
+        let decoder = JSONDecoder()
+        decoder.dateDecodingStrategy = .iso8601withFractional
+        let due = try decoder.decode(DueReview.self, from: json)
+
+        #expect(due.review.cardId == "c1")
+        #expect(due.card.deckId == "d1")
+        #expect(due.card.type == .basic)
+        #expect(due.card.fields["front"] == "Was ist 1+1?")
+    }
+
+    @Test("Rating-Enum-Werte sind exakt wie ts-fsrs erwartet")
+    func ratingValues() {
+        #expect(Rating.again.rawValue == "again")
+        #expect(Rating.hard.rawValue == "hard")
+        #expect(Rating.good.rawValue == "good")
+        #expect(Rating.easy.rawValue == "easy")
+    }
+}