gh-98586: add whatsnew entry covering limited-API outgoing vector calls, related documentation fix

[wjakob]

PR #98587 addressing issue #98586 lacked a "what's new" entry.

While making those changes, I noticed an inconsistency in how PY_VECTORCALL_ARGUMENTS_OFFSET is declared in the underlying Sphinx markup when compared to other macro constants like Py_TPFLAGS_HAVE_VECTORCALL. This PR fixes that as well, which should connect a few currently broken cross-references

• Issue: Add vector call functions to the limited API #98586

[CAM-Gerlach]

FYI, you should mention in the PR title (which is also the default commit message) what you are adding a What's New entry for...

[CAM-Gerlach]

Seems like these should be changed to use equivalent roles from the appropriate domain, though I'm not sure why the Python domain has data/const but the C domain doesn't, despite C actually having "real" constants and Python only having a convention, nor what the reason was for using construct from the "wrong" domain originally...

@AA-Turner , any expert insight here?

[wjakob]

@CAM-Gerlach -- I updated the PR title (sorry) and merged the change with the missing article.

I am hesitant to make the domain-related changes you requested. Part of the reason is that I am following the pattern that generally seems to be used for macro constants. For example look at how Py_TPFLAGS_MANAGED_DICT is defined and referenced in the Sphinx doc (or any macro constant, as far as I can tell). If you wanted to move things over to the :c: Sphinx domain, then it would be good to do this in a concerted set of changes covering the entire documentation.

[CAM-Gerlach]

Okay, that makes sense for now I guess. In that case, LGTM, aside from one comment

[wjakob]

aside from one comment

Incorporated, thanks.

[erlend-aasland]

LGTM, thanks!

Related: we should really update the docs on best practice for writing exension modules using the Limited C API (there is already an issue for this).