Do not let StarPU queue a request on a node which doesn't have any worker

src/datawizard/data_request.c

@@ -184,6 +184,7 @@ void _starpu_post_data_request(struct _starpu_data_request *r, unsigned handling
 {
 	/* We don't have a worker for disk nodes, these should have been posted to a main RAM node */
 	STARPU_ASSERT(starpu_node_get_kind(handling_node) != STARPU_DISK_RAM);
+	STARPU_ASSERT(_starpu_memory_node_get_nworkers(handling_node));
 
 //	_STARPU_DEBUG("POST REQUEST\n");