[evlatests] 20 second subscans in pointing

[Barry Clark]

The usual pointing scans have 10 second subscans, with two subscans
done for each pointing offset position.  (This was done decades
ago when we despaired of getting the flagging good enough to flag
out antenna motion.)  In theory, we have good flagging now, so I
tried organizing pointing with 20 second subscans.  The results are
confusing.

The phenomenology is that for each subscan, there are nine one second
integrations with no fringes and eleven with fringes.  The beginning
of subscan occurs, we think, at the beginning of the last of the nine
fringeless integrations.  The first of the fringing integrations have
higher amplitudes than the remainder on the half-power subscans,
somewhat lower than the others on the on-source point.  But this
may have a natural explanation in terms of antenna motion.  The
systematic effect is the vanishing of fringes 12 seconds into the
subscan, and the extraordinary 9-11 cadence.