Upload & capture

Upload a whole PDF, run the OCR pipeline once, and save the result as a pipeline case expected_output for regression testing.