Can Visual Language Models Replace OCR-Based Visual Question Answering Pipelines in Production? A Case Study in Retail | Read Paper on Bytez