chore: initial project scaffold from architecture design

- Add CLAUDE.md with project overview, tech stack, build commands,
  architecture description, coding standards, and sample images section
- Add full directory structure: src/, docs/, tests/, import/
- Add CMakeLists.txt with C++20, OpenCV/LibRaw/Qt6 dependencies,
  converter_core static lib, optional GUI, and GTest tests
- Add architecture documentation: ARCHITECTURE.md, PIPELINE.md, MODULES.md
- Add source skeletons for all pipeline stages:
  RawLoader, Preprocessor, NegativeDetector, Inverter, ColorCorrector,
  CropProcessor, OutputWriter, Pipeline, MainWindow, CliRunner, main.cpp
- Add initial test stubs for pipeline and rawloader
- Add sample ARW files in import/ for integration testing

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

CLAUDE.md

@@ -0,0 +1,89 @@
+# CLAUDE.md
+
+This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
+
+## Project
+
+C++ desktop app converting digitized analog film negatives (35mm, 120mm) to digital positives. Supports JPG/PNG and RAW input (CR2/NEF/ARW/DNG), outputs PNG/JPG with inversion, C41/B&W correction, auto-crop, and batch processing. Cross-platform: Windows/Linux/macOS.
+
+## Tech Stack
+
+- **Language:** C++20
+- **Image processing:** OpenCV 4.10+ and LibRaw 0.21+ (RAW demosaicing)
+- **GUI:** Qt 6.8 LTS (LGPLv3)
+- **Build:** CMake 3.20+, vcpkg (Windows), Ninja
+- **CLI:** Optional batch mode without GUI
+
+## Build Commands
+
+```bash
+# Linux (Ubuntu/Debian)
+sudo apt install libopencv-dev libqt6widgets6 libraw-dev cmake ninja-build
+cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release
+cmake --build build
+
+# Windows (vcpkg)
+vcpkg install opencv[contrib] libraw qt6-base:x64-windows
+cmake -B build -G Ninja -DCMAKE_TOOLCHAIN_FILE=vcpkg.cmake
+cmake --build build
+
+# macOS (Homebrew)
+brew install opencv libraw qt@6 cmake ninja
+cmake -B build -G Ninja
+cmake --build build
+```
+
+## Architecture
+
+Processing pipeline:
+
+```
+Input → Loader → Preprocess → Detect → Invert → Color Correction → Post-Process → Output
+```
+
+**Core data structure:**
+```cpp
+struct ImageData {
+    cv::Mat rgb;      // Demosaiced 16-bit
+    std::string make; // "Canon", "Nikon"
+    float exposure;   // WB/EXIF data
+};
+```
+
+**Pipeline stages:**
+1. **Loader** (`src/converter/rawloader`): LibRaw for RAW→RGB16, OpenCV for JPG/PNG
+2. **Preprocess**: RAW→RGB16 conversion, deskew
+3. **Detect** (`src/converter/negative`): Negative vs. positive via histogram analysis and orange color masking
+4. **Invert**: `cv::bitwise_not()` + film-specific color matrix
+5. **Color**: C41 orange cast removal, auto white balance from EXIF
+6. **Post** (`src/converter/crop`): Levels, sharpening, dust removal, auto frame detection
+7. **Output**: 16-bit TIFF and 8-bit PNG
+
+Error handling uses `std::expected<ImageData, Error>` throughout.
+
+## Coding Standards
+
+- Always call `LibRaw::recycle()` after use
+- Qt file dialogs: `QFileDialog::getOpenFileNames("RAW (*.cr2 *.nef *.dng)")`
+- Tests use RAW golden files with pixel diff tolerance <1%
+- Do not exceed 4GB in-memory RAW data
+- Do not use lossy demosaicing (LibRaw default is lossless)
+- Always log RAW metadata
+
+## Sample Images
+
+The `import/` directory contains example RAW files for manual testing and development:
+
+| File | Format | Description |
+|------|--------|-------------|
+| `import/DSC09246.ARW` | Sony ARW | Example negative for conversion testing |
+| `import/unbenannt.ARW` | Sony ARW | Example negative for conversion testing |
+
+Use these files to test the full pipeline end-to-end without needing external test data.
+
+## License Compliance
+
+README.md must include:
+- Qt LGPLv3 attribution and source download link: https://www.qt.io/download-open-source
+- Instructions for re-linking Qt DLLs
+- LibRaw CDDL attribution: https://www.libraw.org