Task 04: OCR Pipeline (DeepFellow easyOCR)

[zalun]

Summary

Implement the OCR extraction module that sends PDF/image files to DeepFellow's easyOCR API for deterministic text extraction.

Changes

• src/docproc/ocr.py — Async OCR extraction with retry logic
• tests/test_ocr.py — ~20 test functions
• src/docproc/config.py — Add ocr_endpoint field to DeepfellowConfig
• config-example.yaml — Add ocr_endpoint: "/v1/ocr"
• tests/test_config.py — Add ocr_endpoint to MINIMAL_CONFIG
• pyproject.toml — Version bump, add httpx/pytest-asyncio deps, add asyncio_mode = "auto"
• Version bump 0.1.3 → 0.1.4
• CHANGELOG.md — Add [0.1.4] entry

Acceptance Criteria

• Can extract text from PDF/image files via DeepFellow easyOCR API
• Returns structured OCRResult with page-level breakdown
• Handles API errors with retry logic (exponential backoff)
• Works async for parallel execution with Vision
• Unsupported file types raise clear error
• All quality gates pass (ruff, ty, pytest ≥80% coverage)