TASAC: A twin-actor reinforcement learning framework with a stochastic policy with an application to batch process control