Assessment of Reward Functions for Reinforcement Learning Traffic Signal Control under Real-World Limitations