Fix leap second validation when parsing time_points

[statementreply]

1. Fixed utc_time parsing incorrectly accepting invalid 60th second values (e.g. 00:00:60)
2. Removed leap second validation in sys_time parsing. It now accepts 23:59:59 even if it is deleted by a negative leap second, which can be represented as a sys_time value but doesn't physically exist in UTC.
3. Merged leap second validation and correction code to avoid looking up the leap second table twice.

Fixes #2163. Fixes #2698.

[StephanTLavavej]

I've pushed a conflict-free merge with main, followed by renaming the enum class and dropping _Negative for ABI reasons (see comment). I've verified with dumpbin /symbols and undname that this affects the mangled name, which is good.

[strega-nil-ms]

Other than the nit, this looks good to me.

Requested changes have been made

[StephanTLavavej]

I've pushed a couple of small commits to combine Nicole's clarity improvement (of handling the error case first, before we forget what we tested) with if constexpr {} else {}'s behavior of instantiating only the taken branch, and a comment update because the relative positions of the code are being reversed.

[StephanTLavavej]

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

[StephanTLavavej]

Thanks! That's one small step for a PR, one giant leap second for C++ <chrono>! 🧑‍🚀 ⌚ 😹