[parallelization] Try to get PAM running on our cluster

[MartinPyka]

Try to make use of all cores of our local cluster.

[pherbers]

So I have tried to make a python-only version of our core that can run outside of Blender, and therefore would also be easy to implement for MPI. Problem is, pure python is reeeaaally slow. While technically working, the code in the mesh_representation branch is at least 10 times as slow as the master, and I don’t see any way of getting it up to the same speed in python.

Our options are:

• Leave it running in Blender, stays fast on single systems, will probably not work with MPI
• Try an external library written in a low language with python binding, such as CGAL (Especially the AABB-tree would give us the functionality of the closest point on mesh, all UV methods would have to be reimplemented by hand though)

I'll just try to leave it in Blender for now and just rework the structure to make it more modular in the future.