Integrate get_te0_golomb() calls into the code, this allows some checks to be avoided and the function is pretty small. 3% speedup, though this is probably due to changed inlining and not directly this change.