Bug: a88/881

This would avoid the need to pass around the "use empty correction"
flag for all operations needing the magnetization vector. The value
would be stored per-Datum as with current sample properties, to 
preserve the flat file format. This would also allow different samples
to have different settings -- e.g. if there are two empty slots in
a run (though that might involve complicating the interface confusingly
for little practical benefit).

By extension: should Zplot settings, current correction settings etc.
be handled similarly? At this stage maybe it's worth thinking about
bumping per-sample properties up to the Sample object, even if they
will be flattened out and repeated in every Datum line on saving
(and read from the first one on loading).

Empty-slot corrections deferred to later release; see bug #82
for rationale.