release banzai floyds to pypi

[jchate6]

E2E tests failing in lco-banzai v1.37.1:

https://github.com/LCOGT/banzai-floyds/actions/runs/36498143521/job/109182419317?pr=106

=========================== short test summary info ============================
95
ERROR tests/test_e2e.py::TestWavelengthSolutionCreation::test_if_arc_frames_were_created
96
ERROR tests/test_e2e.py::TestWavelengthSolutionCreation::test_if_wavelength_rmse_is_small

Unit tests failing in lco-banzai v1.38.2:

https://github.com/LCOGT/banzai-floyds/actions/runs/36499506343/job/109186752437?pr=106

=========================== short test summary info ============================
380
FAILED banzai_floyds/tests/test_fringing.py::test_super_fringe_interpolates_pixels_masked_in_every_flat - AttributeError: 'NoneType' object has no attribute 'data'
381
FAILED banzai_floyds/tests/test_fringing.py::test_create_super_fringe - AttributeError: 'NoneType' object has no attribute 'data'

Any chance you two could take a look at what changes might have broken these tests?

[timbeccue]

Looks like these are from some recent changes I made in banzai. I didn't think to verify those against floyds or nres, my bad. These should be pretty simple fixes though, and I'll remember to check these plugins in the future.

One problem is post_to_archive_queue was moved from banzai.utils.file_utils to banzai.utils.messaging. I think this function makes more sense in a messaging module rather than file_utils. If you're fine keeping that there, it's a two line change to fix the import in banzai-floyds (and the same needs to happen in banzai-nres too):

# banzai_floyds/tests/test_e2e.py

-from banzai.utils import file_utils
+from banzai.utils import messaging
 ...
-            file_utils.post_to_archive_queue(frame['filename'], 
+            messaging.post_to_archive_queue(frame['filename'],

The other issue is from a change to init_master_frame that looks for science data in an HDU named 'SCI' rather than the primary HDU. I looked through banzai and thought that everything worked like this, but didn't think outside the repo. Looks like Floyds creates some test frames that have a single unnamed HDU, which propagates into the eventual NoneType error when fringe.py calls init_master_frame. I've reverted back to using primary_hdu in banzai in #496, so once that's merged and tagged, updating floyds to use 1.39.1 should resolve that error.

[timbeccue]

The HDU lookup has been fixed in banzai 1.39.1: https://github.com/LCOGT/banzai/releases/tag/1.39.1