DIFDAT differences adjacent data points in an object, creating
a new dataset with axis and data_array components. The
aim is to reduce low frequency noise in a time series. If a
quality component exists, DIFDAT will ignore the bad points,
warning the user that bad data is present, calculating the
output data value by difference*(mean spacing/gap). DIFDAT
also works on primitive datastets and HDS files with irregular
axes, although in this case DIFDAT is less applicable and the
user is warned that it is irregular.