Better support to check pandas chains

[filipsch]

Up till now, pythonwhat does not bother with 'package specific' support for testing python scripts. The pandas package, however, is extremely commonly used across a wide range of Python courses, as it is the basis package for almost any data analysis with tabular data. Currently, you can test chains of pandas calls with pythowhat, but the support is far from great. There are no elegant ways to allow for both of these solutions, for example:

df.group_by('[b]').mean()
df.group_by('b').mean()

Or for these solutions

df = pd.DataFrame({'a': [1, 2, 3]})

# select the column a
df['a']
df.a

We should carefully think about additional functions we could add to the pythonwhat API to robustly check whether a student did the correct data manipulation operations. I have no idea what it could possibly look like, but the impact could be huge both in terms of CD productivity and SCT quality.