feat(parse): implement OCR text extraction for image parser

[mvanhorn]

Description

Implement the _ocr_extract() method in ImageParser using pytesseract (Python binding for Tesseract OCR). The method was a stub returning None with an explicit TODO at image.py:203.

Follows the same async pattern from _asr_transcribe() in the audio parser (PR #805): wraps the synchronous pytesseract call in asyncio.run_in_executor() and degrades gracefully when pytesseract is not installed.

Related Issue

Relates to #372

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)
• Performance improvement
• Test update

Changes Made

• Replace _ocr_extract() stub in openviking/parse/parsers/media/image.py with pytesseract integration
• Add [ocr] optional dependency group in pyproject.toml (pip install openviking[ocr])
• Add tests in tests/parse/test_image_ocr.py covering: text extraction, empty text, missing pytesseract, exception handling, language parameter passthrough

Testing

• 5 test cases in tests/parse/test_image_ocr.py using mocked pytesseract
• Verified locally with Tesseract 5.5.2 on a generated test image containing "OpenViking OCR Test 2026" - text was extracted correctly
• ruff format and ruff check pass

Why this matters

ImageParser already has VLM description support (_vlm_describe) and config fields for OCR (enable_ocr, ocr_lang in ImageConfig), but _ocr_extract returned None. Images containing text (screenshots, documents, whiteboards) lost their textual content during ingestion. This fills the gap using the same pattern that worked for audio transcription.

Design decisions

• pytesseract over PaddleOCR: lighter dependency (no torch). Chinese text works via chi_sim lang pack in Tesseract.
• Optional dependency: pytesseract is not added to core deps. ImportError returns None with a warning, matching how the codebase handles optional providers.
• No new config fields: reuses existing enable_ocr and ocr_lang from ImageConfig.

This contribution was developed with AI assistance (Claude Code).

[github-actions]

Failed to generate code suggestions for PR

[mvanhorn]

The build distribution CI failures (No module named pip in the isolated build env) appear to be a pre-existing infrastructure issue, not related to the changes in this PR.

The ocr optional dependency group I added to pyproject.toml triggers the check-deps gate (which normally skips the build matrix). The failure happens during python -m build --wheel bootstrapping - before the build tool even reads the new extras. The Linux container build (ubuntu:20.04) loses pip accessibility after the git reset --hard && git clean -fd workspace clean step.

lint, test-lite, check-deps, and CLA all pass. Are the build distribution jobs expected to pass for PRs that touch pyproject.toml, or is this a known issue?