Update the regression test to METAL observed/AST dataset

[karllark]

The current regression tests are based on the beast-examples/phat_small dataset (obs/AST results). The details of the phat_small dataset are lost in the mists of time. The phat_small dataset is old enough that it does not have the full information that we now use in the BEAST (e.g., the ASTs only report mags for the recovered measurements, instead of fluxes that all more recent AST results provide).

This new dataset is based on a subset of a METAL field.

A number of regression tests are skipped as they need updated cache files either for phat or metal or both. See #598.

Includes minor updates to plot_indiv_fit.py.
Includes various updates to close files left open and remove some deprecation warnings. This cleans up the test output making it easier to see problems.

Partially addresses #567.
Closes #325.

[codecov-commenter]

Codecov Report

Merging #612 into master will decrease coverage by 4.23%.
The diff coverage is 18.33%.

@@            Coverage Diff             @@
##           master     #612      +/-   ##
==========================================
- Coverage   47.25%   43.01%   -4.24%     
==========================================
  Files          94       94              
  Lines        9079     9123      +44     
==========================================
- Hits         4290     3924     -366     
- Misses       4789     5199     +410     

Impacted Files	Coverage Δ
beast/plotting/plot_noisemodel.py	12.50% <ø> (ø)
beast/tools/run/create_physicsmodel.py	42.85% <ø> (-23.81%)	⬇️
beast/tools/setup_batch_beast_fit.py	8.73% <ø> (ø)
beast/observationmodel/ast/make_ast_input_list.py	7.54% <2.77%> (-1.40%)	⬇️
beast/tools/run/make_ast_inputs.py	15.51% <6.66%> (-3.24%)	⬇️
beast/physicsmodel/prior_weights_dust.py	100.00% <100.00%> (ø)
beast/plotting/plot_cmd.py	91.42% <100.00%> (+0.25%)	⬆️
beast/plotting/plot_cmd_with_fits.py	94.44% <100.00%> (+0.21%)	⬆️
beast/plotting/plot_indiv_fit.py	95.75% <100.00%> (-0.12%)	⬇️
beast/tools/convert_hdf5_to_fits.py	93.18% <100.00%> (+0.15%)	⬆️
... and 13 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0d26d1a...707495d. Read the comment docs.

[lea-hagen]

Looks good! I like the new file organizational system for the tests (both in the new class and in the folders online).

One comment about plot_indiv_fit: the filters are still hard-coded in, so it'll only work for a particular survey. I'd imagine it would be straightforward to extract them from the stats file, and maybe use pysynphot or something to get the wavelength from the corresponding filter curve. But that may be outside the scope of this PR :).

For the tests, perhaps getting everything consistently regenerated would be a good topic for an upcoming hack day??? 😄

[karllark]

One comment about plot_indiv_fit: the filters are still hard-coded in, so it'll only work for a particular survey. I'd imagine it would be straightforward to extract them from the stats file, and maybe use pysynphot or something to get the wavelength from the corresponding filter curve. But that may be outside the scope of this PR :).

Good point! I've been wondering why plot_indiv_fit has been looking odd for the metal results. Will see what I can do about this in the next PR (not this one).

[karllark]

For the tests, perhaps getting everything consistently regenerated would be a good topic for an upcoming hack day???

Good idea. I've been thinking we should have one soon. I'll setup a doodle poll.

[lea-hagen]

Thanks for making the changes!