Support formatting of time_point with utc_clock

[patrickroocks]

According to the discussion in #3098 in this PR a formatter for std::chrono::time_point<std::chrono::utc_clock> is added. Also added a small unit test. fmtlib must be compiled with C++20 or later to use this, as std::chrono::utc_clock is a C++20 feature.

[vitaut]

Thanks for the PR! Overall looks good, just a few nits inline.

[patrickroocks]

Thanks for the review! Pushed the fixes.

[vitaut]

Looks good but there are some CI failures.

[patrickroocks]

The chrono test fails with

unknown file: error: C++ exception with description "The specified module could not be found." thrown in the test body.
(https://github.com/fmtlib/fmt/actions/runs/3125553362/jobs/5078549521#step:6:134)

As this happens only for MSC version 1929 but not for version 1933 (the newest one), I check for this version now.

[mwinterb]

It's probably not related to the MSVC version. Windows doesn't "historically" use the IANA time zone names, and so to support them for tzdb, Microsoft's STL uses ICU to handle that translation. icu.dll is only included in newer versions of Windows. Windows Server 2019 is based on Windows 1809, so it is too old to have just "icu.dll".

to_sys for utc_time uses tzdb to handle leap seconds.

[phprus]

@patrickroocks, @vitaut,
Maybe disable this test for Windows?
Like here:

fmt/test/chrono-test.cc

Lines 153 to 156 in 48f525d

	// MSVC:
	// minkernel\crts\ucrt\src\appcrt\time\wcsftime.cpp(971) : Assertion failed:
	// timeptr->tm_year >= -1900 && timeptr->tm_year <= 8099
	#ifndef _WIN32

[vitaut]

Maybe disable this test for Windows?

Fine with me.

Also if @mwinterb is correct then the current MSVC check will only mask the issue and should be removed.

[patrickroocks]

Removed the MSVC check, disabled the utc_clock test for windows. Squashed the commits.

[vitaut]

Thank you!