Prevent reclaim in send_traverse_thread()

As is the case with traverse_prefetch_thread(), the deep stacks caused
by traversal require disabling reclaim in the send traverse thread.

Fixes: openzfs#4912

module/zfs/dmu_send.c

@@ -573,6 +573,7 @@ send_traverse_thread(void *arg)
 	struct send_thread_arg *st_arg = arg;
 	int err;
 	struct send_block_record *data;
+	fstrans_cookie_t cookie = spl_fstrans_mark();
 
 	if (st_arg->ds != NULL) {
 		err = traverse_dataset_resume(st_arg->ds,
@@ -585,6 +586,7 @@ send_traverse_thread(void *arg)
 	data = kmem_zalloc(sizeof (*data), KM_SLEEP);
 	data->eos_marker = B_TRUE;
 	bqueue_enqueue(&st_arg->q, data, 1);
+	spl_fstrans_unmark(cookie);
 }
 
 /*