Add OCR support for scanned PDF invoices

[sdamianiw]

Currently ChemTrace only parses digital-native PDFs (text-selectable). Many German SMEs still receive scanned invoices. Scope: integrate an OSS OCR engine (Tesseract or docling) as optional dependency. Priority: post-MVP, dependent on pilot feedback.