More various cleanups

[StephanTLavavej]

Some of these are things I've noticed while reading every line of the STL's headers for Standard Library Modules, but they aren't blocking anything.

• In tests/libcxx/expected_results.txt, change 2 tests from FAIL to SKIPPED. This avoids unexpected passes when running set_environment before running the tests. (Because set_environment puts the freshly-built STL on the INCLUDE path, Clang suppresses warnings from those "system headers", so if tests are expected to FAIL due to such warnings, they will instead unexpectedly pass.)
  • This is a quality-of-life improvement for myself, as I always run set_environment.
• Update .github/workflows/update-status-chart.yml as new versions of actions/checkout and actions/setup-node are now available. Also update node-version (as usual, it's not critical to keep this constantly updated.)
  • Tested in my fork.
• In <variant>, DevCom-1004719 "Bogus warning C4100 'unreferenced formal parameter' emitted by if constexpr" was resolved as fixed, but it still repros in the compiler's own sources. I've filed VSO-1513409 internally with a preprocessed file, so we should update the bug citation.
• Several classes were implicitly using private access control. We conventionally say private explicitly. (With exceptions for friend declarations and static_asserts where the access control doesn't matter.)
  • Initially noticed while reading all headers, then audited via regex, so this list is fairly comprehensive (didn't bother looking outside STL headers).
• We were unnecessarily repeating access specifiers in <sstream> and <xtree>.
  • Noticed while folding class members in VSCode - I don't have a more automated way to scan for this.
• In <charconv>, we should use _STL_INTERNAL_STATIC_ASSERT for our internal array sizes - there's no reason to perform such validation during user compilations. (The static assertions themselves are probably overly paranoid.)
• Comment the <cwctype> #undefs.
  • At first, I thought we could remove them like in <cctype>: do not defend against isalnum/etc. macros #2147, but after I looked at the UCRT, they appear to be needed for /clr:pure 😿, so we should record this for vNext.

[StephanTLavavej]

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