QtWayland doesn't attach a null buffer to a wl_surface to unmap it,
instead it destroys the wl_surface. This is not wrong, but it breaks a
lot of assumptions about the way how the qpa manages window system
resources.

With this change, KWindowShadow will start using heuristics to keep
track of window system resources and when needed, re-create shadow
objects. We need to use heuristics because there is nothing in public
API that we can use.

It's worth noting that QtWayland provides a private API that you could
use to get notified when the wl_surface object has been destroyed. But
we should use private APIs only if there is no other way around. In
this particular case, we have one (QEvent::Expose and QEvent::Hide).