v0.3.0 - Vision Model Support & i18n

Features:
- Multimodal (image-text-to-text) model support with mmproj auto-conversion
- Python-based model conversion for complex architectures (Qwen3.5, etc.)
- Image upload in chat with JPEG compression and thumbnail optimization
- Internationalization (zh/en) with browser language auto-detection
- API Info dialog with vision model image parameter examples
- Chat session clear history and copy feedback

Fixes:
- Fix localStorage quota crash by stripping image data before storage
- Fix llama-server reasoning mode for Qwen3.5
- Sanitize unsupported image formats before sending to llama-server
- Align Library action buttons
- Add cache-control headers for HTML

Scripts:
- Update install scripts for macOS/Linux (.sh) and Windows (.ps1)