deps: drop openai-agents + pytesseract; promote pdf2image, remove OCR leg
- pyproject.toml: drop unused `openai-agents` and `pytesseract` from [project] dependencies
- think/importers/documents.py: promote `pdf2image` from optional try/except to a plain hard import; remove `_extract_text_ocr` and the OCR leg of the scanned-PDF fallback chain; `process()` scanned path is now vision → sparse-pypdf-with-error
- tests/test_importer_documents.py: delete OCR fallback test; collapse the remaining scanned fallback test to single-leg vision-failure behavior
- uv.lock: regenerated