Reinforcement Learning Random Access for Delay-Constrained Heterogeneous Wireless Networks: A Two-User Case.