Major OCR improvements:
* Higher recognition accuracy and stability
* Faster processing on typical documents
* Better multi-language handling
* More consistent, searchable PDF output
* Improved text extraction across diverse layouts

Camera improvements:
* Camera rotation now updates automatically
* Flash detection improved (CameraX API)
* Light sensor tied to lifecycle (saves battery)
* Torch separated from capture flash
* Safer camera re-init on errors
* Better memory cleanup for previews
* Localized error messages
* FileProvider authority centralized
* Low-light prompt improved
* Back navigation refined