Fix SDSS-I/II spSpec units

[aragilar]

It looks like something changed (I've not yet worked out what) in how astropy/wcslib handled absent CUNITa, and now the loader is incorrectly using metres instead of using angstroms. As the WCS was already incorrectly encoded, we probably can't trust CUNITa anyway, so force angstroms.

This also adds a test to confirm that the unit does not change again.

[codecov]

Codecov Report

Merging #1066 (38774cf) into main (a9c432c) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##             main    #1066   +/-   ##
=======================================
  Coverage   70.72%   70.72%           
=======================================
  Files          64       64           
  Lines        4485     4485           
=======================================
  Hits         3172     3172           
  Misses       1313     1313           

Files Changed	Coverage Δ
specutils/io/default_loaders/sdss.py	47.19% <100.00%> (ø)

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[pllim]

"The WCS in the SDSS files does not appear to follow the WCS standard" o_o

Since it is wonky to begin with, I guess this is fine, but please confirm with SDSS that this is universal for all their DRs. If not, you might also need to filter by DR number or something?

[aragilar]

The WCS issue is only for the spSpec files, so only affects SDSS-I and SDSS-II, SDSS-III and later use a different format and hence loader, and are based on FITS tables rather than WCS, and so avoid all the WCS unit issues.

https://classic.sdss.org/dr7/products/spectra/read_spSpec.php and https://classic.sdss.org/dr7/dm/flatFiles/spSpec.php document the format, and appear to be similar for the earlier data releases as well, and are identical for the relevant parts documenting the wavelength scale. These emphasise the misuse (i.e. sticking the log parameters into a linear WCS) is intentional, it may be that the format (or at least the choice of it) predates the standardisation of logarithmic WCS?

I've added links to the above pages to the docstrings, so the next person who looks at that code knows what the format is documented as (rather than having to dig through issues).

[pllim]

Hard to know if this works for sure given failure caused by this:

• TST: SDSS remote data file link is broken #1072

[weaverba137]

This all looks good to me. For historical perspective, the spSpec files were designed to be compatible with spectrum viewers in IRAF, so some differences in format are to be expected.

[pllim]

@aragilar , can you please rebase? The devdeps job should be all green unless this PR introduced new failures.

[pllim]

CI is green. Thanks!

[pllim]

I dunno how this package do milestones, so I'll let @rosteen or @tepickering handle that.