Update on "[ONNX] Update Reducesum operator for opset 13 (#50532)"

* udpate symbolic for squeeze/unsqueeze

* update c++ unsqueeze/squeeze creation

* clang format

* enable tests

* clang format

* remove prints

* remove magic number

* add helper function

* fix build issue

* update opset9 symbolic with helper function

* fix utility test

* fix prim_fallthrough opset skip

* enable reducesum opset 13

* enable embedding_bag which contain reducesum op

* add ReduceSum helper

* remove block_listed_operators

* remove local test code

* remove embedding_bag() in opset13 file

* remove unuse import

Co-authored-by: BowenBao <bowbao@microsoft.com>
Co-authored-by: hwangdeyu <deyhuang@qq.com>

[ghstack-poisoned]

torch/csrc/jit/passes/onnx/shape_type_inference.cpp

@@ -265,8 +265,8 @@ Node* CloneNodeToGraph(
     int opset_version) {
   auto vals_to_params_map =
       buildValueToParamsMap(n->owningGraph()->block(), params_dict);
-  auto clone_node =
-      n_graph->createClone(n, [&n_graph, &vals_to_params_map, opset_version](Value* v) {
+  auto clone_node = n_graph->createClone(
+      n, [&n_graph, &vals_to_params_map, opset_version](Value* v) {
         auto v_n = v->node();
         switch (v_n->kind()) {
           case ::c10::onnx::Constant: {