Restructure: Move ocr_pipeline + labeling + crop into ocr/ package

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

klausur-service/backend/ocr/cv_pipeline.py

@@ -0,0 +1,37 @@
+"""
+CV-based Document Reconstruction Pipeline for Vocabulary Extraction.
+
+Re-export facade — all logic lives in the sub-modules:
+
+  cv_vocab_types      Dataklassen, Konstanten, IPA, Feature-Flags
+  cv_preprocessing    Bild-I/O, Orientierung, Deskew, Dewarp
+  cv_layout           Dokumenttyp, Spalten, Zeilen, Klassifikation
+  cv_ocr_engines      OCR-Engines, Vocab-Postprocessing, Text-Cleaning
+  cv_cell_grid        Cell-Grid (v2 + Legacy), Vocab-Konvertierung
+  cv_review           LLM/Spell Review, Pipeline-Orchestrierung
+
+Lizenz: Apache 2.0 (kommerziell nutzbar)
+DATENSCHUTZ: Alle Verarbeitung erfolgt lokal.
+"""
+
+from .types import *       # noqa: F401,F403
+from .preprocessing.preprocessing import *     # noqa: F401,F403
+from .layout.layout import *            # noqa: F401,F403
+from .engines.engines import *       # noqa: F401,F403
+from .cell_grid.cell_grid import *         # noqa: F401,F403
+from .detect.box_detect import *         # noqa: F401,F403
+from .review.review import *            # noqa: F401,F403
+
+# Private names used by consumers — not covered by wildcard re-exports.
+from .preprocessing.preprocessing import _apply_shear  # noqa: F401
+from .layout.layout import (  # noqa: F401
+    _detect_header_footer_gaps,
+    _detect_sub_columns,
+    _split_broad_columns,
+)
+from .engines.engines import (  # noqa: F401
+    _fix_character_confusion,
+    _fix_phonetic_brackets,
+)
+from .cell_grid.cell_grid import _cells_to_vocab_entries  # noqa: F401
+from .words_first import build_grid_from_words  # noqa: F401