Fix C++20/gcc-12 issues (Part 1)

[falbrechtskirchinger]

This is the first small batch of fixes to address build failures in C++20 mode and/or using gcc-12.

It incorporates 3 commits from @nlohmann, although cc01459 should arguably not be part of this (should I remove that commit?).
The spaceship operator is needed to fully address #3138 and #3207 (Part 2) and CI will not pass all checks.

Fixes #3138.
Closes #3226.

NOTE: Based on my update-ci branch. The (currently) 3 bottom commits are not part of this PR. Once #3368 is merged I can rebase and remove draft status.

[coveralls]

Coverage remained the same at 100.0% when pulling 5c51ac7 on falbrechtskirchinger:topic/fix-c++20-issues-part1 into 4a6e6ca on nlohmann:develop.

[falbrechtskirchinger]

The remaining CI failure is the one observed here: #3138 (comment)
To be fixed in part 2 (I don't have a 100% working version yet and have refocused on improved unit testing to aid debugging).

The changes here are needed for PR #3380 not to reveal even more broken tests.

Do you want me to disable the failing test temporarily? As selectively as possible.
I've disabled the failing comparisons.

[nlohmann]

Thanks for incorporating #3226, and please leave cc01459 in - it's part of the GCC update.

[nlohmann]

Looks good to me.

[nlohmann]

Will merge when the CI is done.

[falbrechtskirchinger]

Will merge when the CI is done.

Great. Feel free to cancel all my unrelated, pending AppVeyor CI runs.

[nlohmann]

Good point - it's a pity AppVeyor is so slow...

@falbrechtskirchinger Thanks so much for your recent contributions!

[falbrechtskirchinger]

@falbrechtskirchinger Thanks so much for your recent contributions!

@nlohmann You're very welcome! There are more on the way, so I'd like to thank you for reviewing them as juggling all these items is becoming mentally taxing.