AttackViz dataset + robust chart QA model from ChartAttack: Testing the Vulnerability of LLMs to Malicious Prompting in Chart Generation