Fix boolalpha extraction to be case-sensitive again

[StephanTLavavej]

Fixes #1541.

_Getloctxt() is a non-member function template. We can change its signature because it's header-only. (I verified with dumpbin /exports that it isn't exported from the DLL. It's called by separately compiled virtual member functions, which is totally fine. As an aside, this means that the __CRTDECL is unnecessary; I will consider a global cleanup later.)

For clarity, I'm introducing enum class _Case_sensitive (as bool arguments are generally unreadable) and making it mandatory (no default argument).

As a targeted cleanup, because I'm already changing the signature (and this is always called with template argument deduction), I'm reordering the template parameters to match the function parameters.

The core fix is a multi-line conditional expression, but due to the naming I believe it's readable. This seemed less invasive than restructuring the control flow. (I also felt that testing this at runtime was okay, compared to templating on _Case_sensitive and using if constexpr.)

Then we need to change all callsites (not searching for these is what led to the regression; a useful lesson). boolalpha must be case-sensitive, but all of the <xloctime> calls are correctly case-insensitive. (Future cleanup: ":AM:am:PM:pm" is an unnecessary relic now that _Getloctxt() handles lowercase, uppercase, and mixed case.)

Finally, I'm adding the test case that I developed, which is expressed with my new favorite technique, static tables.

[mnatsuhara]

Sorry I forgot this one was on my plate! 🍽️ Looks great!