Learning to Continually Learn with the Bayesian Principle