Fixes for failures when using astropy develop.

[WilliamJamieson]

Currently, poppy is causing failures in the jwst development dependencies CI. These failures are due to a change in the some of the APIs in astropy introduced by the merger of PR astropy/astropy#12633. This PR resolves these issues, allowing poppy to properly work with the current state of the astropy develop branch.

[codecov]

Codecov Report

Merging #482 (911b732) into develop (2640d89) will decrease coverage by 1.97%.
The diff coverage is 100.00%.

@@             Coverage Diff             @@
##           develop     #482      +/-   ##
===========================================
- Coverage    74.73%   72.76%   -1.98%     
===========================================
  Files           18       18              
  Lines         6476     6476              
===========================================
- Hits          4840     4712     -128     
- Misses        1636     1764     +128     

Impacted Files	Coverage Δ
poppy/utils.py	53.53% <100.00%> (ø)
poppy/matrixDFT.py	63.63% <0.00%> (-34.66%)	⬇️
poppy/instrument.py	55.68% <0.00%> (-11.61%)	⬇️
poppy/geometry.py	90.90% <0.00%> (-1.14%)	⬇️
poppy/fresnel.py	83.59% <0.00%> (-0.98%)	⬇️
poppy/accel_math.py	37.09% <0.00%> (-0.60%)	⬇️
poppy/wfe.py	80.81% <0.00%> (-0.43%)	⬇️
poppy/poppy_core.py	80.55% <0.00%> (-0.36%)	⬇️
poppy/dms.py	45.85% <0.00%> (-0.28%)	⬇️
poppy/optics.py	81.78% <0.00%> (-0.12%)	⬇️

---

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 2640d89...911b732. Read the comment docs.

[nden]

@mperrin @shanosborne I am not sure who should review this PR.
Could you assign someone?
We'd appreciate a quick resolution and a release, as we are about to make a jwst release in a week.

[mperrin]

@nden Will do! @shanosborne are you able to review this for a quick turnaround?

[WilliamJamieson]

The test failures are likely due to the numpy 1.22 changes, see numpy/numpy#19478. These changes can cause minor numerical, which may require updating the tolerances used for np.allclose tests.

It maybe possible to disable some cpu features instead, see astropy/astropy#12684 for details.

[nden]

Ping to get attention to this issue.
For what it's worth it appears I have permissions to review it so I did.

[mperrin]

Hi @nden and @WilliamJamieson - thanks for the ping. @shanosborne said she's available to take a look at this today. (She was out last week; thanks for your patience)

[shanosborne]

This change fixes the astropy develop issue and the test_rotations failures in the CI are unrelated to this PR so I can fix them separately and not hold this PR up. This is good to go.