ENH #1854: Tests will check for alternate baselines using basename_[1…

[danlipsa]

…-9].ext

Previously the pattern used was baseline.*.ext

[danlipsa]

CDAT/uvcdat-testdata#127

[aashish24]

LGTM 👍

[danlipsa]

@doutriaux1 Please review.

[doutriaux1]

@danlipsa I'm cool with your changes, but in general I think it's unsafe to use alternate baselines.

First it adds work to maintaining the baselines, also I'm always worried a test gets better doesn't need the laternate baseline anymore, then we fix an issue with that test. A few week later a new change breaks the test but we don't see it because the old "bad" baseline is still here and passes.

[danlipsa]

@doutriaux1 I am not happy with alternate baselines but I think they are a necessary evil. It boils down to the following choice: Should we spend a week to debug a small difference in how images are rendered on a specific platform, possibly narrow it down to a OpenGL difference in how the standard is interpreted or a OpenGL bug -- or spend a minute to add a new baseline. VTK uses alternate baselines. We run on many platforms/compilers. I think we only use 3 alternate baselines at most. For uvcdat I think we have at most one.

[doutriaux1]

agreed