A Two-stage Framework and Reinforcement Learning-based Optimization Algorithms for Complex Scheduling Problems