A Day-to-Day Dynamical Approach to the Most Likely User Equilibrium Problem