test: matplotlib backend side-effects

[jodom961]

As described in the issues https://github.com/ydataai/pandas-profiling/issues/888 and https://github.com/ydataai/pandas-profiling/issues/837 , upon the release of 3.1.0, a side effect of importing pandas_profiling was that it overrode the existing backend of matplotlib on import, and prevented plots from being rendered in Jupyter environments.

Since it seems the backend does not needed to be set per matplots documentation,

By default, Matplotlib should automatically select a default backend which allows both interactive work and plotting from scripts, with output to the screen and/or to a file, so at least initially you will not need to worry about the backend.

Using matplot.libuse() will require changes in your code if users want to use a different backend. Therefore, you should avoid explicitly calling use unless absolutely necessary.

and it is generally a best practice to not have side effects upon importing a package, I have elected to remove that line.

A new test is in place in notebooks/notebook_plotting.ipynb to prevent this in the future.

[jodom961]

@sbrugman clean start for tests. let me know if it needs anything else to be approved.

[sbrugman]

Thanks @jodom961. The comments above provide some pointers to improve the PR.