Make angle-weighted interpolation for topology optimization available for parallel computing

[sblauth]

At the moment, topology optimization does not work in parallel.

[sblauth]

The issue arises due to the angle-weighted interpolation. Most likely this is the case due to an indexing error as the mesh and dofs are distributed among multiple tasks.

[sblauth]

This is not so easy to do - it requires setting parameters["ghost_mode"] = "shared_vertex" from the fenics side and then some careful manipulation of the owned vertices of each process (to ensure that things are not done multiple times). I think for now it is easier to raise an error if topology optimization is run with interpolation_scheme = angle and it is detected that the program runs on more than one MPI process.