Learning the Limits: How System Limitation Transparency Shapes Mental Models, Trust, and Acceptance in Conditionally Automated Driving Systems