This basically reverts #3766.

Reason being that screenX/Y doesn't work when locking the pointer. Its also not ideal because when you move the pointer out of the frame and enter it again, it pretends like the mouse moved a lot, which is quite different then what movementX/Y reports and probably not what we want.

Seeing that its unclear in the spec how browsers should implement movementX/Y I also opened w3c/pointerlock#100.

Now that we don't have a choice but use movementX/Y we have to account for the following:

1. Pretend like Unclear if pointermove should set movementX/Y w3c/pointerlock#100 has been resolved and all browsers actually set values there instead of zero. Which they currently do.
2. Account for browsers not actually following the spec which coordinate system to use. See movementX/Y coordinate space w3c/pointerlock#42.
3. Account for Firefox actually being buggy when un-coalescing events. See Bugzilla.

To account for all these issues, I added some browser engine detection code, which I might expose to the user in a follow-up PR.

Also have to update #2875 after this is merged.