Update pointer focus before motion occurs

[ids1024]

This seems to work, but produces a flood of events. Smithay probably need a method to update focus without producing events unless a surface is entered.

Instead of time: 0, this and other synthesized events should probably use CLOCK_MONOTONIC. It seems libinput does document that it uses that.

[ids1024]

Actually, we can just test PointerHandle::current_focus() to see if we need to change it. So maybe no smithay changes are needed.

[Drakulix]

This seems to work, but produces a flood of events.

Why does it flood though. refresh_focus is called on an idle-callback, right? So it should only happen once all events are processed and the compositor would go into sleep, no?

[ids1024]

My earlier draft didn't have the pointer.current_focus() check so it sent motion events with the same position every time the idle callback was run. Now this seems to be reasonable.

[ids1024]

I think this should be correct. The use of current_location().as_global() seems to match what is used elsewhere, and this now doesn't seem to produce input events unnecessarily.

I think the timestamp here and elsewhere shouldn't be 0, but if so that can be updated in various places after Smithay/smithay#1509 is merged.