Strengthen tests in 10mins notebook

[MattHJensen]

In #1603 I asked:

I do wonder why our notebook tests didn't catch this though.

@martinholmer replied:

Because the notebook tests are very weak. They only detect a fatal Python error.
They have never done any validity checking on the answers the notebooks produce.

All I'm saying is that many of the results produced in the Behavioral_example.ipynb notebook are numerically incorrect after the merge of PR #1582. The notebook runs without a Python crash, but produces many incorrect results.

As @martinholmer further pointed out, this indicates a need for stronger tests in the 10min notebook.

[martinholmer]

This comment illustrates the weakness of the testing logic in the test_notebooks.py file.

[MaxGhenis]

I'm not familiar with the issue here, but would be happy to contribute example notebooks as a form of documentation if it can be resolved. I have a variety in https://github.com/MaxGhenis/taxcalc-notebooks of varying degrees of complexity, ranging from notebook versions of @martinholmer's recipes to larger analyses.

[martinholmer]

Over the past several months, this issue has been resolved indirectly by developing a Tax-Calculator Cookbook of tested Python scripts that use the taxcalc package. These scripts are tested before each release and can be easily uploaded into Jupyter Notebooks by users who like the interactive nature of notebooks.

[martinholmer]

Given the development of the Cookbook, this issue is being closed.