Adaptive cubic regularisation methods for unconstrained optimization. Part I: motivation, convergence and numerical results