test_perfect_model_prediction takes a while to run

[ahuang11]

I haven't looked into it; but maybe could drop redundant tests or optimize the function it's testing.

[aaronspring]

agree. the 3D tests takes really long. maybe we test all in 1D and just a few selected metrics in 3D. or just subselect 3x3 grids

[bradyrx]

We're already subselecting a small grid in 3D in test_perfect_model_prediction. I imagine a lot of the time is spent on loading in the datasets too? Maybe we just create 3D dummy data here..

[ahuang11]

When I have time I'll try profiling the code and try to optimize it

[aaronspring]

when subselecting 3D we should maybe use xr.open_mfdataset or xr.open_dataset(files, chunks={}) to only load the subselected data?

[aaronspring]

probably we dont have to test whether all metrics pass the bootstrap function, but rather testing one metric for all different setups and once testing all metrics.