feat(phase1): implement OpenCV image processing pipeline

- Create ImageProcessor service with EXIF orientation detection
- Implement smart cropping via OpenCV contour detection (10% padding)
- Add text orientation detection using Hough line transform
- Resize and compress images to 1200px with 85% JPEG quality
- Generate 200px square thumbnails with center crop
- Fallback to Pillow if OpenCV fails
- Comprehensive test suite: 28 tests all passing
- File size validation (reject >10MB)
- Graceful error handling for corrupted/invalid images
- Update requirements.txt with opencv-python, piexif, python-magic

venv/lib/python3.12/site-packages/cv2/flann/__init__.pyi

@@ -0,0 +1,65 @@
+__all__: list[str] = []
+
+import cv2
+import cv2.typing
+import os
+import typing as _typing
+
+
+# Enumerations
+FLANN_INDEX_TYPE_8U: int
+FLANN_INDEX_TYPE_8S: int
+FLANN_INDEX_TYPE_16U: int
+FLANN_INDEX_TYPE_16S: int
+FLANN_INDEX_TYPE_32S: int
+FLANN_INDEX_TYPE_32F: int
+FLANN_INDEX_TYPE_64F: int
+FLANN_INDEX_TYPE_STRING: int
+FLANN_INDEX_TYPE_BOOL: int
+FLANN_INDEX_TYPE_ALGORITHM: int
+LAST_VALUE_FLANN_INDEX_TYPE: int
+FlannIndexType = int
+"""One of [FLANN_INDEX_TYPE_8U, FLANN_INDEX_TYPE_8S, FLANN_INDEX_TYPE_16U, FLANN_INDEX_TYPE_16S, FLANN_INDEX_TYPE_32S, FLANN_INDEX_TYPE_32F, FLANN_INDEX_TYPE_64F, FLANN_INDEX_TYPE_STRING, FLANN_INDEX_TYPE_BOOL, FLANN_INDEX_TYPE_ALGORITHM, LAST_VALUE_FLANN_INDEX_TYPE]"""
+
+
+
+# Classes
+class Index:
+    # Functions
+    @_typing.overload
+    def __init__(self) -> None: ...
+    @_typing.overload
+    def __init__(self, features: cv2.typing.MatLike, params: cv2.typing.IndexParams, distType: int = ...) -> None: ...
+    @_typing.overload
+    def __init__(self, features: cv2.UMat, params: cv2.typing.IndexParams, distType: int = ...) -> None: ...
+
+    @_typing.overload
+    def build(self, features: cv2.typing.MatLike, params: cv2.typing.IndexParams, distType: int = ...) -> None: ...
+    @_typing.overload
+    def build(self, features: cv2.UMat, params: cv2.typing.IndexParams, distType: int = ...) -> None: ...
+
+    @_typing.overload
+    def knnSearch(self, query: cv2.typing.MatLike, knn: int, indices: cv2.typing.MatLike | None = ..., dists: cv2.typing.MatLike | None = ..., params: cv2.typing.SearchParams = ...) -> tuple[cv2.typing.MatLike, cv2.typing.MatLike]: ...
+    @_typing.overload
+    def knnSearch(self, query: cv2.UMat, knn: int, indices: cv2.UMat | None = ..., dists: cv2.UMat | None = ..., params: cv2.typing.SearchParams = ...) -> tuple[cv2.UMat, cv2.UMat]: ...
+
+    @_typing.overload
+    def radiusSearch(self, query: cv2.typing.MatLike, radius: float, maxResults: int, indices: cv2.typing.MatLike | None = ..., dists: cv2.typing.MatLike | None = ..., params: cv2.typing.SearchParams = ...) -> tuple[int, cv2.typing.MatLike, cv2.typing.MatLike]: ...
+    @_typing.overload
+    def radiusSearch(self, query: cv2.UMat, radius: float, maxResults: int, indices: cv2.UMat | None = ..., dists: cv2.UMat | None = ..., params: cv2.typing.SearchParams = ...) -> tuple[int, cv2.UMat, cv2.UMat]: ...
+
+    def save(self, filename: str | os.PathLike[str]) -> None: ...
+
+    @_typing.overload
+    def load(self, features: cv2.typing.MatLike, filename: str | os.PathLike[str]) -> bool: ...
+    @_typing.overload
+    def load(self, features: cv2.UMat, filename: str | os.PathLike[str]) -> bool: ...
+
+    def release(self) -> None: ...
+
+    def getDistance(self) -> int: ...
+
+    def getAlgorithm(self) -> int: ...
+
+
+