Simultaneous Training of First- and Second-Order Optimizers in Population-Based Reinforcement Learning