REDA: A Novel AI Approach to Multi-Agent Reinforcement Learning That Makes Complex Sequence-Dependent Assignment Problems Solvable