gpu_api: add dummy gpuStream{Create,Destroy,Synchronize} for host

One could actually think about having an actual implementation using
threads, but this will be good enough for now.

src/fortran/gpu_api.cxx

@@ -184,4 +184,16 @@ extern "C" int gpuMemcpyAsync(void* dst, const void* src, size_t bytes,
   return 0;
 }
 
+#elif defined(GTENSOR_DEVICE_HOST)
+
+// dummy implementation, one GPU only
+extern "C" int gpuStreamCreate(gt::stream_view::stream_t* pStream) { return 0; }
+
+extern "C" int gpuStreamDestroy(gt::stream_view::stream_t stream) { return 0; }
+
+extern "C" int gpuStreamSynchronize(gt::stream_view::stream_t stream)
+{
+  return 0;
+}
+
 #endif