Envoyer par SMS: A Multi-Agent Reinforcement Learning Approach to Price and Comfort Optimization in HVAC-Systems