Full implementation - Vision AI, config, improved pipeline

Major changes:
- vision_analyzer.py: Real OpenAI/Anthropic vision API integration
  - Component detection with confidence scores
  - Atomizer hints extraction (objectives, constraints, parameters)
  - Material and feature identification
  - Timeline correlation with transcript

- config.py: Full configuration system
  - API settings (provider, keys, models)
  - Processing settings (Whisper model, frame interval, scene detection)
  - Output settings (BOM, hints, PDF template)
  - Config file support (~/.cad-documenter.toml)

- audio_analyzer.py: Enhanced transcription
  - Audio stream detection
  - Graceful fallback for missing audio
  - Keyword extraction
  - Technical term detection
  - Timeline correlation

- video_processor.py: Smart frame extraction
  - Scene change detection via ffmpeg
  - Configurable thresholds
  - Best frame selection

- doc_generator.py: Improved output
  - Better Markdown templates
  - BOM CSV export
  - Atomizer hints JSON
  - Component cards

- cli.py: Rich CLI with progress indicators
  - Config file support
  - --init-config flag
  - Verbose mode
  - Better error messages

- tests: Comprehensive test suite

pyproject.toml

@@ -15,6 +15,8 @@ dependencies = [
     "jinja2>=3.1.0",
     "openai-whisper>=20231117",
     "pillow>=10.0.0",
+    "httpx>=0.27.0",
+    "tomli>=2.0.0;python_version<'3.11'",
 ]
 
 [project.optional-dependencies]