A 24x Speedup for Reinforcement Learning with RLlib + Ray

[@Mohsenghq]

Thanks for explaining simple, I migrated to RLlib and it's really efficient.