<cmath>, <limits>: Locally disable -Wnan-infinity-disabled for Clang

[frederick-vs-ja]

Fixes #4931.

Possibly because of that numeric_limits<FP> are implemeted as full specializations, when one attempts use min() or epsilon(), Clang emits warning for Nan and Inf related functions, even if these functions are not themselves called.

This PR suppresses -Wnan-infinity-disabled in <limits> and <cmath> locally to avoid such false positives. When a user explicitly calls infinity(), quiet_NaN(), or signaling_NaN(), the warning is still triggered (Godbolt link).

[CaseyCarter]

Suppressing these warnings at template parse time while allowing them at instantiation is an excellent solution. I didn't know this was possible before.

[StephanTLavavej]

Thanks! 😸 ♾️ @CaseyCarter I pushed a macro refactoring after you approved, moving them to <yvals_core.h> below our usual Clang suppression macros, using the same escape-hatch pattern (just in case), and recording the warnings that they're suppressing in comments.

[StephanTLavavej]

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

[StephanTLavavej]

♾️ 🚫 #️⃣