Now we can know the position of each node/polygon/point of the animated model before it's even drawn on screen.
We can also know the position of the points on screen (and the depth of those points).
The dirty box of the models is now computed (in a debug fashion).
Somehow, models head are still bugged but it's probably due to the fact that they are not animated (and not initialised).
The debug display is left activated to let Endy see the results and implement a better zbuffer refresh system
