Double number format only from now on

The entire library now uses double as a number format. That has the advantage that then that the transforms assuming real valued time series really have double and not some typedef.