Remove {UN,}SET_S4_OBJECT usage

[MichaelChirico]

Part of #6180.

I would like to test the dogroups change. I don't think there's any reason to do UNSET_S4_OBJECT()? How could the object wound up marked as S4?

Certainly we pass tests as is. Can anyone come up with an example broken by this change? 🤔

[github-actions]

Generated via commit 801267c

Download link for the artifact containing the test results: ↓ atime-results.zip

Time taken to finish the standard R installation steps: 11 minutes and 46 seconds

Time taken to run atime::atime_pkg on the tests: 3 minutes and 17 seconds

[ben-schwen]

AFAIU IS_S4_OBJECT is also part of non exported API.

[MichaelChirico]

I don't think that's right, here it is in Rinternals:

https://github.com/r-devel/r-svn/blob/53bbd920926a6263bf0e870fa142c6171f37ec01/src/include/Rinternals.h#L273

Well above the "Not API" section

And here it's mentioned in R-ints as the "correct" way to test S4:

https://github.com/r-devel/r-svn/blob/53bbd920926a6263bf0e870fa142c6171f37ec01/doc/manual/R-ints.texi#L1397-L1401

[ben-schwen]

My bad, I should have linked the new section in WRE.

There we have it:

This table lists some non-API functions used in packages and the API functions that should be used instead:
IS_S4_OBJECT
Use isObject and isS4.

[MichaelChirico]

Nice. Slightly confused why this is note showing up in our NOTE (yet?), and about the organization of the header & R-ints ref, but assume that'll be ironed out over time.

[ben-schwen]

LGTM, but lets see what revdeps will say