Use the new pyuvdata analytic short dipole beam in our polarized source test.

[bhazelton]

Description

Replaces the ad hoc analytic short dipole used in our testing of polarized sources with the pyuvdata analytic short dipole beam.

This requires bumping the minimum pyuvdata version to v3.1.0.

I also bumped the minimum lunarsky version to 0.2.5 for compatibility with astropy 6.0

Motivation and Context

Checklists:

Other Change Checklist

• I have read the contribution guide.
• My code follows the code style of this project.
• Any new or updated docstrings use the numpy docstring format.
• I have updated the tutorial to highlight my new feature (if appropriate).
• I have added tests to cover any changes.
• My change includes a breaking change
  • My change includes backwards compatibility and deprecation warnings (if possible).
• I have updated the CHANGELOG if appropriate.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 100.00%. Comparing base (eb75e59) to head (2f9fecc).
Report is 20 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main      #268   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            5         5           
  Lines         2121      2120    -1     
  Branches       543       543           
=========================================
- Hits          2121      2120    -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jpober]

I'm happy to approve this, but the fact that we need to transpose, flip, and flip the beam to get the pyuvdata analytic beam to match pyuvsim seems notable. I assume it's a massive overhaul to make pyuvsim use the pyuvdata beam convention? This is pyradiosky, not pyuvsim, so it shouldn't be a hold up on this PR -- but having different beam conventions floating around seems bad...

[bhazelton]

I'm happy to approve this, but the fact that we need to transpose, flip, and flip the beam to get the pyuvdata analytic beam to match pyuvsim seems notable. I assume it's a massive overhaul to make pyuvsim use the pyuvdata beam convention? This is pyradiosky, not pyuvsim, so it shouldn't be a hold up on this PR -- but having different beam conventions floating around seems bad...

This is done internally in the visibility creation code in pyuvsim (in uvsim.py in the apply_beam method on the UVEngine object). It's not an externally available beam representation at all. The reason to do it here is so that I can just copy the code that creates the visibility from pyuvsim. I'm not entirely sure I understand why all this re-arrangement is done in pyuvsim, I think it's something to do with how the einsum is set up. If we want to fix it, we should do it in pyuvsim.

[jpober]

Looks great then. I'll merge.