Fix seldest2src(), not even wrong.
Use specks_reupdate() after changing thresh, so the reported
selcount should be right.
specks_rethresh(): keep track of whether anything changed,
and increment st->selseq if so.