Compare average code length with source entropy to compute redundancy (avgLen - H) / H; lower redundancy means a better code.
冗余越低编码越优,理想情况趋近 0。