Consider automatically defining LIBDIVIDE_SSE2 et al

[AlexGuteniev]

For Visual Studio:

• /arch compiler option sets target processor and enables compiler to use corresponding instructions
• /arch:IA32, /arch:SSE, /arch:SSE2, /arch:AVX, /arch:AVX2, /arch:AVX512 options available
• x86 is by default /arch:SSE2
• x64 is always at least /arch:SSE2, it cannot be /arch:SSE or /arch:IA32
• The value passed to arch is exposed as _M_IX86_FP predefined macro (distinguish SSE2), __AVX__, __AVX2__, and others. See Predefined macros

So libdivide could have something like:

#if defined(_MSC_VER) && defined(__AVX2__) 
#define LIBDIVIDE_AVX2

Maybe other compilers also expose compilation against specific instruction set via preprocessor

[kimwalisch]

We already discussed in https://github.com/ridiculousfish/libdivide/issues/34 why we don't enable SSE2 by default.