Human-Agent Trust Calibration in Multi-Agent AI Deployments: Complacency, Rejection, and Evidence-Based Oversight