C++20 support for constexpr and SIMD simultaneously

[PazerOP]

With C++20 and std::is_constant_evaluated(), it becomes possible to provide both simd and constexpr:

namespace detail
{
	constexpr bool is_constant_evaluated()
	{
#if __cpp_lib_is_constant_evaluated >= 201811
		return std::is_constant_evaluated();
#else
		// The user's compiler doesn't support std::is_constant_evaluated(). 
		// This should be a user-configurable switch to decide if they want
		// SIMD or constexpr (since they have to choose one or the other).
		return true;
#endif
	}
}

constexpr glm::vec4 some_function()
{
	if (detail::is_constant_evaluated())
	{ 
		// standard c++ implementation
	}
	else
	{
		// SIMD implementation, you can do anything here, intrinsics, reinterpret_cast, etc
	}
}

In the above example, if the user's compiler supports std::is_constant_evaluated(), this works exactly as expected:

	constexpr auto slow_but_compile_time_version = some_function();
	const auto fast_runtime_version = some_function();

On older compilers, depending on the value of the user-configurable switch, fast_runtime_version would become "slow" version, or slow_but_compile_time_version would not compile. This is the current behavior, forcing the user to decide between constexpr and SIMD.

[sharkautarch]

@PazerOP
I’ve made a draft pr that implements support for using simd and constexpr simultaneously: #1313

note that you have to set this define before including glm headers for it to work:
#define GLM_SIMD_CONSTEXPR 1
Also three other things to note:

• this pr makes use of gcc’s vector extensions for the simd part of the constructors, so it’s only compatible with gcc and clang
• Only constexpr constructor support is added, the operator overloads don’t currently work at compile time
• Directly using swizzling expressions when calling c-style va_arg functions is broken (ex: printf("%f\n", v.y); always prints 0.0000). C++-style variadic template functions should work fine tho <- fixed

UPDATE: I've now added constexpr support to the operators as well.
I think my draft PR is definitely getting closer to being able to become a real PR, I think I just need to clean up the code a bit more...