Sobolev space is wrong for discontinuous elements

[garth-wells]

The Sobolev space is wrong for (nearly) all elements that are discontinuous. The Basix code doesn't modify the space when the element is discontinuous.

I don't know why the Sobolev space and the continuous/discontinuous parameters are both required. Seems redundant.

[garth-wells]

Fix in #638 for Lagrange elements. Other elements still need to be fixed.