Better use of std::enable_if

[beinhaerter]

Replace all occurances of class = std::enable_if_t<Cond> and typename = std::enable_if_t<Cond> with std::enable_if_t<Cond, bool> = true.

This commit is inspired by #1174, which changed one occurance in the owner header. This commit is aimed to fix all remaining occurances.

[carsonRadtke]

I had originally considered asking for this change as part of #1174, but I was worried there may be subtle differences that need to be considered for each change.

From my point of view, I would like justification for each change (i.e. an example of accepted code that shouldn't be accepted). If the std::enable_if_t is impossible to override, I don't think the change is necessary. That being said, I will bring this PR up during the monthly maintainers sync (which happens to be this evening) to get a second opinion.

[beinhaerter]

I had originally considered asking for this change as part of #1174, but I was worried there may be subtle differences that need to be considered for each change.

From my point of view, I would like justification for each change (i.e. an example of accepted code that shouldn't be accepted). If the std::enable_if_t is impossible to override, I don't think the change is necessary. That being said, I will bring this PR up during the monthly maintainers sync (which happens to be this evening) to get a second opinion.

Sounds good. I am not a template expert. I just saw the original PR and player around a bit myself. I have not tested if every changed occurance was wrong/bad.

[beinhaerter]

I did some investigation. It seems not be possible to explicitely choose the template instantiation for a templated constructor in a class template, so my changes for not_null and span were not useful, I removed them.

For the other changes I split my PR into three commits:

1. add the tests that should fail but that do not fail
2. do the code changes that make the tests fail as expected
3. revert all changes from the first commit

I would have preferred to keep the tests from the first commits, but my SFINAE skills are too low. I tried it the way span_compatibility_tests is doing it, but I did not get it working.

[beinhaerter]

I am now happy with the PR. I was able to write tests like static_assert(XXCompilesFor<YY>, ""); that work for the code before the fix and where negating the condition like static_assert(!XXCompilesFor<YY>, ""); works after the fix. So we can now keep the tests in place and get rid of the third commit that deleted the tests.

[carsonRadtke]

Thanks for adding examples and it is awesome that they are now part of the tests. I will take another look once the tests are green for all compilers.

FYI - I made a change in the repo settings; the actions should now automatically run for your PRs.

[carsonRadtke]

This looks great! Thanks for making all those changes.