Handle beam tables with >1 stokes component: fix for #825

[e-koch]

This PR implements fixes #825.

• [x] Fix for FITS files, where the beam table is not split into the different components
• [x] Fix for CASA images where we don't have reading beams for different components implemented.
• [x] Tests for reading/writing Stokes cubes with beam tables

[codecov-commenter]

Codecov Report

Merging #826 (00b32c2) into master (b4ff491) will increase coverage by 0.25%. The diff coverage is 85.71%.

@@            Coverage Diff             @@
##           master     #826      +/-   ##
==========================================
+ Coverage   77.93%   78.19%   +0.25%     
==========================================
  Files          24       24              
  Lines        5861     5920      +59     
==========================================
+ Hits         4568     4629      +61     
+ Misses       1293     1291       -2     

Impacted Files	Coverage Δ
spectral_cube/io/casa_image.py	66.05% <0.00%> (ø)
spectral_cube/conftest.py	93.49% <100.00%> (+0.72%)	:arrow_up:
spectral_cube/cube_utils.py	83.07% <100.00%> (+0.79%)	:arrow_up:
spectral_cube/io/fits.py	86.62% <100.00%> (+0.17%)	:arrow_up:
spectral_cube/spectral_cube.py	76.58% <0.00%> (+0.03%)	:arrow_up:
spectral_cube/io/core.py	90.90% <0.00%> (+1.51%)	:arrow_up:

---

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 b4ff491...00b32c2. Read the comment docs.

[keflavich]

LGTM. A test would probably be good to add.

Can we merge this with just FITS and move the 'todo' list to issue #825 (and reopen it)?

[e-koch]

I'd like to take a look on how much work the CASA image reader. It may not be too bad and, if so, would be nice to clear off in this PR. I'll update here in a few days.

[e-koch]

The CASA reader fix wasn't too bad.

In both cases we'll still hit issues when the components aren't IQUV, but that's a separate issue.

[e-koch]

Eventually we may want to move some of this into radio-beam for the polarization handling (especially any masking operations across polarizations).