Optional OCR for scanned PDFs

[Phantom-VK]

Problem

The follow-on to issue #54 (scanned PDF silent failure). Users with image-only PDFs need a path forward.
Create sub issues if required, cause dumping all file changes and reviewing them at once , not a good practice.

Requirements

• Must stay opt-in : don't OCR every PDF by default
• Fully local : no cloud APIs, to hold the privacy claim
• Shows progress : OCR on a 200-page scan is slow enough to need a progress bar and cancel button
• User warning : OCR quality varies; document the limitations

Implementation

• Use Tesseract (open source, no network) or any other new lightweight OCR framework available. Do research.
• Add an "Extract text via OCR" button in the File Parser
• Reuse the existing job/progress system (jobs.py, ProgressReport)
• On completion, re-analyze the extracted text

Scope

Involved because it needs:

1. Tesseract detection and error handling
2. A new bridge method in api.py to call OCR
3. Progress tracking in the UI
4. Docs on accuracy and limitations