Bug: a88/b89

If the empty slot correction is applied when no empty slot has
been set, Puffin bombs out with a NullPointerException.

Can't reproduce this; think Christian may be using an old version,
and the bug has since been fixed. Waiting for confirmation of this from C.

Still can't reproduce on my data; waiting for C to send his. From the
backtrace, looks as though a datum isn't getting its sample field set:
the exception is raised at
` if (d.getSample().isEmptySlot()) return d; '
(line 36 of Line.java)

Nasty one, this. Cause: PP ignores ARM and MS lines on loading -- 
they don't even get added to a sample, they're just left for the
garbage collector to clear up after creation. At least that's what's
meant to happen. What actually happened is that Datum's constructor
inserts them into a Line object (introduced for the empty-slot 
correction). When the empty-slot correction is activated, Datum.getUc()
checks for each Datum in its Line whether its sample is in an empty
slot. But if any Datums are ARM or MS, they don't have an associated
sample, and PP bombs out with a null pointer exception.

Ideally, it should be made impossible to construct a Datum without a
Sample, but it's inconvenient in the current implementation, because
the Sample is created after the Datum. As it was, I fixed it by
getting Datum not to add itself to a Line if it's an ARM or MS 
measurement.