Use __builtin_memcpy to implement bit_cast and for accessing vector from pointer of scalars (#53)

* reworking vector_type

* use __builtin_memcpy for bit_cast and vector access of scalar pointer

* clean up

[ROCm/composable_kernel commit: 64350affc5]

host/driver_offline/src/conv_fwd_driver_offline.cpp

@@ -258,7 +258,7 @@ int main(int argc, char* argv[])
     using in_data_t   = half_t;
     using acc_data_t  = float;
     using out_data_t  = half_t;
-#elif 1
+#elif 0
     using in_data_t  = ushort;
     using acc_data_t = float;
     using out_data_t = ushort;