RS-DRL: Managing Uncertainty in Self-Adaptive Systems Based on a Novel Continuous Deep Reinforcement Learning Method