Initial version of multi-step language-based parallelisation exercise

[Traubert]

This is a new exercise discussed in the course planning meeting. You fetch text, preprocess it, run a topic modeling package on it, and try to parallelise the code in four separate places. I couldn't decide which directory to place it in, so I made a new one for it, but it's fine to put it somewhere else instead. I replaced github-for-starter-code with downloading a zip package from Allas.