An optimal adaptive algorithm for the approximation of concave functions