[evlatests] WIDAR oddities

[Vivek Dhawan]

The current data are in C array, the earlier high-dynamic range
observations were in A-array. Michael & I came up with a way to
check more carefully if the shortest baselines have the worst
closure errors, e.g. from cross talk or shadowing or aliasing
at low fringe-rates, or whatever.

The answer is that non-closing errors in the WIDAR0 data are
_not_ dependent on baseline length.

Procedure:

. Take final data set used for imaging (after all self-cal).

. Apply BLCAL with e.g. 10min solint, then SPLIT to make a
   new dataset.

. DIFUV will divide 2 datasets and make a new one, amp~1, Ph~0,
   which one can plot with the familiar UVPLT or VPLOT.

   (All this is needed because there is no way to directly plot
   the contents of a BL solution table, which I hope Eric can
   fix with his usual magic.)


Vivek.