I did some invesigation about the blocked active-changed signal that was
emitted while active-changed was running.
Basically there are two problems:
1) in gramps_main the function change_active_person emits active-changed for
every mouse click, including multiple clicks on the same person in the
People View. change_active_person should only emit this when there actually
was a change.
2) The recursive emission was triggered by the new PedView, that after
building up the gtk tables requested a gtk_main_iteration to get the redraw
done and adopt to the size of the tree. What happened on double clicking was
that the gtk_main_iteration inside the PedView triggered
change_active_person for the second mouse click while the action of the
first was still running.
The PedView now uses idle_add to do the size adaption deferred.