Bezier RefFE dofs

[santiagobadia]

@pmartorell @fverdugo

I think that the Bezier implementation we have in Gridap is not complete yet (if you want to use it for PDE discretisation).

From what I can see here, it is using the same DOFs as the Lagrangian FE. This is not correct, I think.

I open the issue to remember that problem, if I have some time I will try to find the way to define this DOF basis for Bezier.