Workflow restructuring

[rettigl]

As discussed:

• run all-versions tests only on push to main (pull-request merges)
• speed up tests by omitting multiple calculations of inverse dfields in metadata tests

[rettigl]

Disadvantage is: We apparently don't get the coveralls reports if we don't do pull-request builds...

[coveralls]

Pull Request Test Coverage Report for Build 6483750068

• 4 of 4 (100.0%) changed or added relevant lines in 1 file are covered.
• 5 unchanged lines in 2 files lost coverage.
• Overall coverage decreased (-0.1%) to 90.479%

---

Files with Coverage Reduction	New Missed Lines	%
sed/calibrator/energy.py	2	91.86%
sed/binning/numba_bin.py	3	87.74%

Totals
Change from base Build 6471550709:	-0.1%
Covered Lines:	4286
Relevant Lines:	4737

---

💛 - Coveralls

[zain-sohail]

Looks good. Definitely faster now.
I will look into caching the installation of the package as that is mostly same. Would save another 2 mins.
And you said the other slowdown was downloading data for tests right? Maybe we can also cache that, or use a smaller set?

[rettigl]

Looks good. Definitely faster now. I will look into caching the installation of the package as that is mostly same. Would save another 2 mins. And you said the other slowdown was downloading data for tests right? Maybe we can also cache that, or use a smaller set?

That was related to the notebook conversion. I suppose we don't do this often.

[zain-sohail]

Looks good. Definitely faster now. I will look into caching the installation of the package as that is mostly same. Would save another 2 mins. And you said the other slowdown was downloading data for tests right? Maybe we can also cache that, or use a smaller set?

That was related to the notebook conversion. I suppose we don't do this often.

Right. I am not sure if the github workers can actually process the notebooks. I will try that and see. Otherwise, we would have to do it locally or our own runner (possible to set this up now).

I will merge this PR and update a few workflows afterwards.