Text this: Nonlinear adaptation of established linear and predictive control laws through safe reinforcement learning