Mending Trust in AI: Trust Repair Policy Interventions for Large Language Models in Visual Data Journalism