NXP backend: Update the pass move_auxiliary_operator_into_separate_qdq_cluster. (pytorch#13841)

### Summary
Update the passes in `move_auxiliary_operator_into_separate_qdq_cluster_pass.py` to only
transform specific combinations of operators.

### Test plan
Tested by existing tests in `test_edge_passes.py`.

Author: MartinPavella

backends/nxp/edge_passes/move_auxiliary_operator_into_separate_qdq_cluster_pass.py

@@ -11,6 +11,11 @@
 from torch.fx import Node
 from torch.fx.passes.infra.pass_base import PassResult
 
+# Operator aliases for better readability.
+AddMM = exir_ops.edge.aten.addmm.default
+ViewCopy = exir_ops.edge.aten.view_copy.default
+MM = exir_ops.edge.aten.mm.default
+
 
 def insert_qdq_pair_after_node(
     graph: torch.fx.Graph, anchor: torch.fx.Node, q_params: tuple
@@ -41,15 +46,17 @@ def insert_qdq_pair_after_node(
 
 def _is_dequantize(node_: Node) -> bool:
     return (
-        node_.op == "call_function"
+        hasattr(node_, "op")
+        and node_.op == "call_function"
         and node_.target
         == exir_ops.edge.quantized_decomposed.dequantize_per_tensor.default
     )
 
 
 def _is_quantize(node_: Node) -> bool:
     return (
-        node_.op == "call_function"
+        hasattr(node_, "op")
+        and node_.op == "call_function"
         and node_.target
         == exir_ops.edge.quantized_decomposed.quantize_per_tensor.default
     )
@@ -82,20 +89,19 @@ class MoveLeadingAuxiliaryOperatorIntoSeparateQDQClusterPass(NeutronEdgePass):
                                                            ▼
     """
 
-    allowed_auxiliary_nodes = [exir_ops.edge.aten.view_copy.default]
-
-    # List of approved nodes to which the <aux_node> can be connected in order for the pass to make the modification.
-    allowed_main_cluster_nodes = [
-        exir_ops.edge.aten.addmm.default,
-        exir_ops.edge.aten.mm.default,
-    ]
+    # Dictionary mapping main cluster nodes to auxiliary nodes, for which this optimization will be applied.
+    main_cluster_node_to_auxiliary_nodes = {
+        AddMM: [
+            ViewCopy,
+        ],
+        MM: [
+            ViewCopy,
+        ],
+    }
 
     def run(self, graph_module: torch.fx.GraphModule) -> PassResult:
         for aux_node in graph_module.graph.nodes:
-            if (
-                aux_node.op != "call_function"
-                or aux_node.target not in self.allowed_auxiliary_nodes
-            ):
+            if aux_node.op != "call_function":
                 continue
 
             dequantize_node = aux_node.args[0]
@@ -109,11 +115,13 @@ def run(self, graph_module: torch.fx.GraphModule) -> PassResult:
                 continue
 
             main_cluster_node = users[0]
-            if (
-                main_cluster_node.op != "call_function"
-                or main_cluster_node.target not in self.allowed_main_cluster_nodes
+            if main_cluster_node.op != "call_function":
+                continue
+
+            if aux_node.target not in self.main_cluster_node_to_auxiliary_nodes.get(
+                main_cluster_node.target, []
             ):
-                # Unsupported `main_cluster_node`.
+                # Unsupported main cluster node and auxiliary node pair.
                 continue
 
             # Make sure the nodes are part of the same QDQ cluster.
@@ -163,29 +171,33 @@ class MoveTrailingAuxiliaryOperatorIntoSeparateQDQClusterPass(NeutronEdgePass):
                                                             ▼
     """
 
-    allowed_auxiliary_nodes = [exir_ops.edge.aten.view_copy.default]
-
-    # List of approved nodes to which the `<aux_node>` can be connected in order for the pass to make the modification.
-    allowed_main_cluster_nodes = [
-        exir_ops.edge.aten.addmm.default,
-        exir_ops.edge.aten.mm.default,
-    ]
+    # Dictionary mapping main cluster nodes to auxiliary nodes, for which this optimization will be applied.
+    main_cluster_node_to_auxiliary_nodes = {
+        AddMM: [
+            ViewCopy,
+        ],
+        MM: [
+            ViewCopy,
+        ],
+    }
 
     def run(self, graph_module: torch.fx.GraphModule) -> PassResult:
 
         for aux_node in graph_module.graph.nodes:
-            if (
-                aux_node.op != "call_function"
-                or aux_node.target not in self.allowed_auxiliary_nodes
-            ):
+            if aux_node.op != "call_function":
                 continue
 
             main_cluster_node = aux_node.args[0]
-            if (
-                main_cluster_node.op != "call_function"
-                or main_cluster_node.target not in self.allowed_main_cluster_nodes
+            if not (
+                hasattr(main_cluster_node, "op")
+                and main_cluster_node.op == "call_function"
+            ):
+                continue
+
+            if aux_node.target not in self.main_cluster_node_to_auxiliary_nodes.get(
+                main_cluster_node.target, []
             ):
-                # Unsupported `main_cluster_node`.
+                # Unsupported main cluster node and auxiliary node pair.
                 continue
 
             users = list(aux_node.users.keys())