Implement P2291R3 constexpr Integral <charconv>

[AdamBucior]

Fixes #2920.

[cpplearner]

I tried to just extend that test but it didn't work well due to constexpr steps limit

😈 Should we increase the limit?

[AdamBucior]

I tried to just extend that test but it didn't work well due to constexpr steps limit

😈 Should we increase the limit?

Unfortunately clang-cl does not recognize that option 🙁

[CaseyCarter]

Just a comment nitpick that I'll apply.

[StephanTLavavej]

Thanks, this looks great! 😻 I pushed small changes for an unnecessary cast and tiny style improvements (FYI @CaseyCarter @strega-nil-ms).

The test's use of internal machinery (_CONSTEXPR23, _Is_constant_evaluated()) is reasonable - there is precedent (we have a few test uses of _CONSTEXPR20), and they'll be easy to address if in the moderate-far future we add Modules configurations to all tests (as import std; will not provide such machinery).

[frederick-vs-ja]

I guess we can avoid using internal mechanisms in test file by replacing them with the following... (no change requested)

Suggested change

	#if _HAS_CXX23 // testing constexpr overloads for integral types in C++23 (P2291R3)
	#define CONSTEXPR23 constexpr
	constexpr is_runtime_evaluated() noexcept {
	return !is_constant_evaluated();
	}
	#else // ^^^ _HAS_CXX23 / vvv !_HAS_CXX23
	#define CONSTEXPR23 inline
	constexpr is_runtime_evaluated() noexcept {
	return true;
	}
	#endif // _HAS_CXX23

[StephanTLavavej]

Yes, we could do something like that in the future, possibly centralized in a lightweight classic header.

For the time being, tests don't need to strenuously avoid internal machinery and macros as long as there is reasonable justification.

[StephanTLavavej]

I'm mirroring this to the MSVC-internal repo - please notify me if any further changes are pushed.

[StephanTLavavej]

Thanks for making <charconv> even more powerful! 🚀 🧮 😻