plotnine incompatibility with scipy

[anupshah14]

Check for the scipy compatibility before the start of the workshop.

[pansapiens]

For the record, we were caught be a reported bug in plotnine that hadn't yet made it to the official release: https://github.com/has2k1/plotnine/issues/275

The current fix is:

!pip uninstall -y scipy
!pip install scipy==1.2.0

# ... then restart the jupyter/colaboratory kernel

import * from plotnine  # should work after kernel restart

We may be able to avoid this in the future by strictly locking all versions with a pip freeze >requirements.txt (or a conda environment.yml) and having participants install this at the start. It would be nice to work in the idea if reproducible environments if it could be kept lightweight. It's less transparent than explicit install commands though, and so might be seen as unneeded complexity (having to explain a requirements.txt vs 'normal' pip install command).