Safe Model-Based Off-Policy Reinforcement Learning for Eco-Driving in Connected and Automated Hybrid Electric Vehicles