Does It Run and Is That Enough? Revisiting Text-to-Chart Generation with a Multi-Agent Approach