Encode offset values into a signed integer to reduce byte cost without too much loss
float or double value to be compressed
Encode offset values into a signed integer to reduce byte cost without too much loss