Can Learned Optimization Make RL Less Difficult?