SVDE: Scalable Value-Decomposition Exploration for Cooperative Multi-Agent Reinforcement Learning