Online Optimisation for Online Learning and Control – From No-Regret to Generalised Error Convergence