Minimizing Error and VLSI Complexity in the Multiplication-Free Approximation of Arithmetic Coding