decode: make quantization step size conform to spec

remove extra factor of 0.5, which should be applied during
the actual quantization step. Note: plugin still uses old step
size, which includes 0.5