[evlatests] A new UVW problem

[Walter Brisken]

All modcomp-free data is afflicted by a problem recently discovered by 
Neal Miller.  Negative valued U, V, and W baseline coordinates were 
subjected to a bit-packing problem that results in an error up to about 
100 meters.  This means that imaging more than a few 10s of synthesized 
beams away from the phase center will be noticably degraded.  UVFIX 
remedies the situation.

A fix for this exists and has been tested, but due to other pending 
changes in the EVLA software this fix is not likely to be installed until 
sometime mid or late next week.

We are discussing if and how to mend the archive.  In the meantime, all 
EVLA data should be UVFIXed...

-Walter