Things got broken. Again. We no longer kept track of the selected dives in
our structures which broke statistics.
This attempts to fix that, but appears to still have a bug when selecting
trips. Sometimes this results in 0 dives being selected according to our
data structures, while Qt happily shows all dives of the trip as seected.
Signed-off-by: Dirk Hohndel <dirk@hohndel.org>