Enabling Rapid Shared Human-AI Mental Model Alignment via the After-Action Review