Select random samples of each group

* This was suggested by David Matheson (from SPSS) (to select 100 from each group).

compute ran1 = uniform(1).
rank variables = ran1 by group / rank into ran2.
select if (ran2 le 100).
execute.