Improve undo/redo test flakyness

[Janpot]

Can't really replicate the flakyness locally but I think we should only start the timeout after we've verified that the page has processed the change. The update only being processed on the page when we're already 450ms into the timeout could explain the flakyness I'm seeing in some test results

[apedroferreira]

If it makes the test less flaky this makes sense.
I also don't remember exactly why we're still debouncing on undo/redo, we should probably be able to get rid of it since the DOM changes are "atomic".