$\mathcal{RL}_1$-$\mathcal{GP}$: Safe Simultaneous Learning and Control