feat: Orientierung + Zuschneiden als Schritte 1-2 in OCR-Pipeline

Zwei neue Wizard-Schritte vor Begradigung:
- Step 1: Orientierungserkennung (0/90/180/270° via Tesseract OSD)
- Step 2: Seitenrand-Erkennung und Zuschnitt (Scannerraender entfernen)

Backend:
- orientation_crop_api.py: POST /orientation, POST /crop, POST /crop/skip
- page_crop.py: detect_and_crop_page() mit Format-Erkennung (A4/A5/Letter)
- Session-Store: orientation_result, crop_result Felder
- Pipeline nutzt zugeschnittenes Bild fuer Deskew/Dewarp

Frontend:
- StepOrientation.tsx: Upload + Auto-Orientierung + Vorher/Nachher
- StepCrop.tsx: Auto-Crop + Format-Badge + Ueberspringen-Option
- Pipeline-Stepper: 10 Schritte (war 8)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

klausur-service/backend/main.py

@@ -42,7 +42,8 @@ try:
 except ImportError:
     trocr_router = None
 from vocab_worksheet_api import router as vocab_router, set_db_pool as set_vocab_db_pool, _init_vocab_table, _load_all_sessions, DATABASE_URL as VOCAB_DATABASE_URL
-from ocr_pipeline_api import router as ocr_pipeline_router
+from ocr_pipeline_api import router as ocr_pipeline_router, _cache as ocr_pipeline_cache
+from orientation_crop_api import router as orientation_crop_router, set_cache_ref as set_orientation_crop_cache
 from ocr_pipeline_session_store import init_ocr_pipeline_tables
 try:
     from handwriting_htr_api import router as htr_router
@@ -177,6 +178,8 @@ if trocr_router:
     app.include_router(trocr_router)      # TrOCR Handwriting OCR
 app.include_router(vocab_router)      # Vocabulary Worksheet Generator
 app.include_router(ocr_pipeline_router)  # OCR Pipeline (step-by-step)
+set_orientation_crop_cache(ocr_pipeline_cache)
+app.include_router(orientation_crop_router)  # OCR Pipeline: Orientation + Crop
 if htr_router:
     app.include_router(htr_router)            # Handwriting HTR (Klausur)
 if dsfa_rag_router: