Don't use _Float16 on non-SSE2 x86

_Float16 on x86 requires SSE2 instruction set to be available.
References:

 * https://gcc.gnu.org/onlinedocs/gcc/Floating-Types.html
 * https://clang.llvm.org/docs/LanguageExtensions.html#half-precision-floating-point

simde/simde-f16.h

@@ -76,7 +76,7 @@ SIMDE_BEGIN_DECLS_
       defined(SIMDE_X86_AVX512FP16_NATIVE) || \
       (defined(SIMDE_ARCH_X86_SSE2) && HEDLEY_GCC_VERSION_CHECK(12,0,0)) || \
       (defined(SIMDE_ARCH_AARCH64) && HEDLEY_GCC_VERSION_CHECK(7,0,0) && !defined(__cplusplus)) || \
-      ((defined(SIMDE_ARCH_X86) || defined(SIMDE_ARCH_AMD64)) && SIMDE_DETECT_CLANG_VERSION_CHECK(15,0,0)) || \
+      ((defined(SIMDE_ARCH_X86_SSE2) || defined(SIMDE_ARCH_AMD64)) && SIMDE_DETECT_CLANG_VERSION_CHECK(15,0,0)) || \
       (!(defined(SIMDE_ARCH_X86) || defined(SIMDE_ARCH_AMD64)) && SIMDE_DETECT_CLANG_VERSION_CHECK(6,0,0))) || \
       defined(SIMDE_ARCH_RISCV_ZVFH)
     /* We haven't found a better way to detect this.  It seems like defining