Convergence Rates for Learning Control Algorithms-A Repetitive Process Theory Approach