[BUG] Flaky unittest "Cannot safely cast non-equivalent float to int64"

[benfred]

Our unittests frequently fail for unrelated reasons to the PR. We should make sure our tests run reliably.

• ValueError: Only one class present in y_true. ROC AUC score is not defined in that case
• tests/unit/test_io.py::test_parquet_lists (assert [[1], [1, 4], [2, 0], [3]] == [[1], [1, 4], [2, 3], [3]])
  http://10.20.13.93:8080/job/nvtabular_tests/1681/console
  http://10.20.13.93:8080/job/nvtabular_tests/1678/console
• TypeError: float() argument must be a string or a number, not 'NoneType’:
  http://10.20.13.93:8080/job/nvtabular_tests/1087/console
  http://10.20.13.93:8080/job/nvtabular_tests/1117/console
• TypeError: Cannot safely cast non-equivalent float to int64
  http://10.20.13.93:8080/job/nvtabular_tests/1109/console
  http://10.20.13.93:8080/job/nvtabular_tests/1150/console

[benfred]

The Only one class present in y_true. ROC AUC score is not defined in that case test failure should be fixed by #564

[benfred]

#689 seems to fix the TypeError: float() argument must be a string or a number, not 'NoneType’: in fillmedian

[benfred]

tests seem stable now