FIX: Convert array-like inputs in eyetracking Calibration

[ishaaaan17]

summary

the docstring of Calibration says positions, offsets and gaze are array-like, but if you passed python lists or tuples they stayed as lists. because of this, cal.plot() crashed on assert isinstance(self["positions"], np.ndarray).

changes

• in Calibration.__init__, convert positions, offsets and gaze to numpy arrays with np.asarray(..., dtype=float) when they are not None.
• in Calibration.plot(), replaced the bare assertion with a ValueError that checks positions and gaze are 2d arrays.
• added list-input parametrization and a check that plot() runs, in test_create_calibration.

[scott-huberty]

ooof. @ishaaaan17 the test you edited was written by me a long time ago! I took this opportunity to make it a little cleaner:

1. make it more DRY. Now we don't duplicate the existing pytest parametrization (which contains all of the kwargs for Calibration), just to turn pass in a few lists instead of arrays.
   2 ) Moves the Calibration kwargs out of the pytest.param decorator and into the function itself. I found that passing 10+ kwargs by parametrization was hard to read!

[ishaaaan17]

haha, no worries at all! the refactored test looks much cleaner and easier to read with the kwargs_template dictionary, instead of that big parametrization tuple with 10+ arguments. thanks for polishing it up!

also i checked the CircleCI build_docs failure, and it looks like an upstream issue from commit 4eee1b3 (mailmap (#14411)). in that commit .mailmap fixed the spelling from Kambiz Tabavi to Kambiz Tavabi, but doc/changes/names.inc:243 still has .. _Kambiz Tabavi: https://github.com/ktavabi.

once that line in names.inc is changed to .. _Kambiz Tavabi:, the doc credit generator should pass.

[scott-huberty]

0ca09ee

OK I Updated the name in 0ca09ee let's see if that does it.

[ishaaaan17]

thanks a lot for the review, the test refactor and the merge @scott-huberty! i really enjoyed working on this with you.