COMP: Constrain x86_64 optimization flags according to CMAKE_SYSTEM_PROCESSOR

[thewtex]

CMAKE_SYSTEM_PROCESSOR defines the target compilation architecture. When
we want x86_64 optimizations, e.g. -march=corei7, only add these flags
when building for this architecture. On Linux/macOS (GCC/AppleClang),
CMAKE_SYSTEM_PROCESSOR is "x86_64" and on Windows (Visual Studio) it is
"AMD64".

[seanm]

Introspecting the system processor is antithetical to cross compiling. It'd be nice to see less of it, not more of it. As you know, I just got rid of a problematic case the other day: #2450

It's still setting -march=corei7, which, in my view, as I've said before, is undesirable:

• If you're building to deploy to the public, you may want to support older processors, which many people out there have
• If you're building for yourself only, you might want an even higher setting, like -march=core-avx2

It seems to be a choice that is best for no one, and best left to the person building the code.

[thewtex]

Introspecting the system processor is antithetical to cross compiling. It'd be nice to see less of it, not more of it. As you know, I just got rid of a problematic case the other day: #2450

Note that CMAKE_SYSTEM_PROCESSOR is not the host system processor. This improves cross compilation support -- it is not antithetical to cross compliing.

It seems to be a choice that is best for no one, and best left to the person building the code.

We do want the person building the code to be able to choose their optimization flags, but we also want good, sensible defaults. Enabling SIMD support on x86_64 is a good default.