Drop pp and shogun nodes

[otizonaizit]

• pp, aka parallel python is a library for parallelization of Python code, that can be used in mdp.parallel as an optional backend. From the website and the downloads it seems that some version of that library supports Python 3. Unfortunately the version available on PyPI is Python 2 only. It is available on Debian/Ubuntu in the package python-pp. But this is package is going to be dropped soon, because Debian and Ubuntu are removing all Python 2 pacakges. There is no Python 3 compatible version on Debian/Ubuntu. It seems the only way to have pp on Python 3 is to download manually from the project webpage, but the Python 3 version is not even up to date with the last Python 2 pp version, and it dates back to 2014.
  I think we should drop it and remove the corresponding compatibility code.

• shogun is a library for machine learning and SVM. It does not seem to be available anymore on PyPI, nor on Debian/Ubuntu. It seems that shogun can be installed via conda-forge or by adding a PPA on Ubuntu. Either we find a way to integrate this in travis, or we should drop the nodes using shogun.