Model-based reinforcement learning for active flow control: escalating performance and accelerated training | Synapse