Fix Bug When Importing Roundtrip

[michaelmckinsey1]

Currently, Hatchet prints a TypeError in each Jupyter cell due to a missing positional argument in manage_jupter_change() in hatchet/external/roundtrip/roundtrip/manager.py. By adding and optional positional argument to the function, we avoid the error when this line is ran.

Additionally, the import statements are re-arranged to avoid importing the roundtrip module unintentionally, which is happening in the below image.

[michaelmckinsey1]

@cscully-allison I made a change here to the import, because roundtrip was being imported before checking the IPython.__version__. I think that is part of why the error happens.

[ilumsden]

If this is an IPython API issue, then yes.

By switching the order, the error should not occur at all.

[cscully-allison]

These look good to me. Not sure why the unit tests are failing, but I couldn't see any side effects this would case.

[ilumsden]

This looks good to me. I have one minor suggestion to make the event handler that you changed more flexible and resilient to future IPython API changes, but that suggestion is not strictly needed at this time.

Also, regarding the CI failing, the cause has to do with the fact that Python 3.7 is EOL. Since flake8 is created by the Python Software Foundation, they frequently change the code so that it's only guaranteed to work with the oldest non-EOL version of Python or above. As a result, flake8 now makes use of the "walrus operator" (i.e., :=) internally, which will cause errors when run under Python 3.7. I've fixed this CI error in #129 by bumping the Python version that we use for black and flake8 checks to 3.9. You can either rebase on top of #129 or copy the relevant changes.

[ilumsden]

@michaelmckinsey1 please rebase on develop. @pearce8 just merged #129, so you should be able to pull in the CI fixes by rebasing.