don't use _umul128, fallback to generic in MSYS2/MINGW32 (#1453)

Author: jalius

include/glaze/util/atoi.hpp

@@ -170,7 +170,7 @@ namespace glz::detail
       // But MinGW on ARM64 doesn't have native support for 64-bit multiplications
       answer.high = __umulh(a, b);
       answer.low = a * b;
-#elif defined(GLZ_FASTFLOAT_32BIT) || (defined(_WIN64) && !defined(__clang__))
+#elif defined(GLZ_FASTFLOAT_32BIT) || (defined(_WIN64) && !defined(__clang__) && !defined(__MINGW32__))
       answer.low = _umul128(a, b, &answer.high); // _umul128 not available on ARM64
 #elif defined(GLZ_FASTFLOAT_64BIT) && defined(__SIZEOF_INT128__)
       __uint128_t r = ((__uint128_t)a) * b;