Apparently some instruments still have a DAS/DAQ/DAE that sometimes
creates a negative time-of-flight in the raw data file. This
demonstrated that the quadratic equation was not correctly converting to
d-spacing when DIFA==0. This PR adds an extra trap to simplify the
calculation in this case. Also changes the order of special cases to
move calculating intermediate results until as late as possible. In
principle, this should execute the case of difa==0 faster as zero is not
squared then square rooted.