[daip] weights, dividing by models

[Eric Greisen]

I suspect that the comments somewhere are not correct.  The weights are
scaled by the square of the amplitudes at present - I suspect that the
amplitude was used in the earliest of days.  Note that this weighting
seems to cause model fits (e.g. CALIB) to fail when the model
improves!  This is why such tasks have a DOWEIGHT parameter to take
weights to some power (e.g. 1/2) so that weights do not differ by so
much between short and long spacings.

Eric Greisen