Apply light preprocessing by default (grayscale, denoise, contrast=5) to improve OCR readability

Consolidate segmentation improvements and code cleanup

Rolling out modular v2

Integrate image segmentation and language detection modules

Enhance document viewer UI: Focus on images, reduce clutter

Improve language detection with mistral-ocr-latest model

Fix deprecated use_column_width parameter, replace with use_container_width

Update UI components based on feedback: remove performance mode, make preprocessing sections collapsible, reduce title size, decenter Mistral AI credit

Reconcile main with preview-improvements branch, implementing modular structure, raw text editing, and enhanced framing