core/macros: avoid using SIGNOF for DIV_ROUND

[kfessel]

using SIGNOF triggers a compiler error see #18849

Contribution description

this removes SIGNOF from DIV_ROUND and DIV_ROUND_UP macro

the default implementation (_Generic will not use it) of SIGNOF would yield wrong values for values > ULLONG_MAX/2

Testing procedure

run the tests

Issues/PRs references

#18849

[benpicco]

What I don't like about this is that it introduces an additional division for those cases where the code is not compile-time const.

[riot-ci]

Murdock results

✔️ PASSED

7411d61 core/macros: avoid using SIGNOF for DIV_ROUND

Success	Failures	Total	Runtime
2000	0	2000	06m:22s

Artifacts

• Documentation preview

This only reflects a subset of all builds from https://ci-prod.riot-os.org. Please refer to https://ci.riot-os.org for a complete build for now.

[kfessel]

in many cases "div" and "mod" are calculated by the same call (if a buildin div is used that the buildin divmod is called once starting if -O2

and if the signof part is required the old variant is also not compact

https://godbolt.org/z/cojafofn4

e.g.: compiler option

-mthumb -march=armv7-m -O2 << hardware div
-mthumb -O2 << no hardware div

DIV_ROUND1(a, b) is this PR
DIV_ROUND2(a, b) is this master for signed types
DIV_ROUND3(a, b) is this master for unsigned types

[kfessel]

closing this one since it provides a different and wrong result:

-15 / -5: master: 2 (wrong), this pr: 4 (wrong), correct: 3 see (#18856)
there are further examples where this pr and or master is wrong
see https://godbolt.org/z/h84nKzKv3

closing this in favour of #18856