Temporal-Logic-Based Reward Shaping for Continuing Reinforcement Learning Tasks