PSPP now implements proper pivot tables

[Ben Pfaff]

Until now, the PSPP output engine has had a pretty basic notion of what
constitutes a table.  It was essentially the same sort of model you'd
find in a word processor, where a table has some number of rows and
columns and you can put what you want in them, join or split them, add
lines between them, and so on.  It's flexible enough, but not well
suited for filling with data.

Over the last few months, I've reimplemented the output engine in terms
of pivot tables, similar to what you'd find in spreadsheet software or
in SPSS.  This is maybe slightly less flexible overall, but it's a lot
easier to deal with as a developer (it takes less code and less thought
to produce output that is at least as nice) and ultimately should allow
for a much better user interface.  Tonight, I got that work to a point
where I was satisfied and pushed it to master.

For now, users probably won't notice big changes.  For some procedures,
the output will appear a little different.  Please report any bugs you
find.

I am now working toward the next step, in which PSPPIRE becomes able to
read SPV files.  After that, I might work to make it able to write them
as well.  It's also now a lot more feasible to add some interface for
editing tables (and for pivoting them!) to the UI.