RL Optimization Cuts Step Time 3.5x for Long Prompts or High Rollouts

@PandaAshwinee Shouldn't matter what type of RL as long as either the prompt is long or instead your rollout n is quite high you can easily compute the potential savings by feeding 3 numbers into this: n_rollout x (prompt_len + response_len) / (prompt_len + n_rollout x response_len)