How is it possible to get optimal infiltration fronts during chemical vapor infiltration with thermal gradients ?