A Numerical Approach Based on the Reproducing Kernel Hilbert Space for Solving a Class of Boundary Value Optimal Control Problems