outlier_detection discrepancy causing JS failure

[JoshMock]

This machine_learning/data_frame_evaluate.yml assertion expects an auc_roc value of 0.99995 but my JS tests consistently return 0.5.

I'm not very familiar with the ML APIs, so the test is a bit hard to follow and see where some language-specific discrepancies might arise. My initial assumption was that it might be a JS floating point precision issue, but I don't see anything obvious that would lead to that.