This appears to have been fixed in the `develop` branch; the line you reference (which is line 870 in 'develop') is now preceded by subtraction of a floating point number:
869 d2 = data - float(offset)
870 d2 *= scale
This will automatically promote
d2 to a floating point type which will have no issue with
*= assignment.