Regression test failures

[karllark]

We have updated things in the BEAST that are causing some of the regression tests to fail. The cached files used by these tests need updating. Most of the standard cached files (i.e., generated by running the phat_small example) have already been updated. The remaining regression tests that need updated files or revising so that they are independent of the cached files are:

Updates for cached files/data needed (would also be useful to add code to the appropriate beast-example subdir to recreate these when needed)

• [ ] test_simobs
• [ ] test_compare_spec_type_inFOV
• [ ] test_compare_spec_type_notFOV
• [ ] test_star_type_probability_no_Av
• [ ] test_calc_depth_from_completeness
• [ ] test_create_physicsmodel_with_subgrid
• [ ] test_create_obsmodel_no_subgrid
• [ ] test_create_obsmodel_with_subgrid
• [ ] test_ast_pick_models
• [ ] test_read_lnp_data
• [ ] test_read_noise_data
• [ ] test_create_obsmodel_with_subgrid
• [ ] test_ast_pick_models
• [ ] test_star_type_probability_all_params

We need to document how to generate the cached test files so that we can maintain these regression tests.

[lea-hagen]

I'm pretty sure the underlying reason for the failures of these is the number of PDF bins used in the fitting. Going from 50 to 100 bins changes the 1D/2D PDFs (obviously) as well as the stats. I keep getting mixed up with how many bins the tests should be using, so it would help to clarify that.

[karllark]

I don't think it is the number of bins, but that the SED file has changed due to removing the LogL < -9. points from the isochrones.

[lea-hagen]

Oh yeah, that too!

[karllark]

These tests are marked to be skipped in PR #596. When each test is fixed, the skip notation should be removed. Skipping the tests allows for the coverage report to be generated which is useful.