python/monarch: pytorch InputBuffer::add() signature changed (meta-pytorch#1739)

Summary:
Pull Request resolved: meta-pytorch#1739

pytorch `InputBuffer::add` API changed. adapt `monarch/python/monarch/_gradient_generator.cpp`
```
  realInputBuffer(node).add(
      input_nr,
      check_and_reduce(node->node, input_nr, std::move(t)),
      std::nullopt,
      std::nullopt);  // Remove: node->node
```

that gradient computation still works correctly after this change needs to be tested.

Reviewed By: vidhyav

Differential Revision: D86025607

fbshipit-source-id: 43599cd13db98da9095baac21f27971ccb8ef0f0

Author: shayne-fletcher

python/monarch/gradient/_gradient_generator.cpp

@@ -420,12 +420,20 @@ struct GradientGenerator {
     DEBUG_PRINT(
         "// add: " << node->node->name()
                    << ", input_nr=" << static_cast<int>(input_nr) << "\n");
+#if !defined(FBCODE_BUILD)
+    realInputBuffer(node).add(
+        input_nr,
+        check_and_reduce(node->node, input_nr, std::move(t)),
+        std::nullopt,
+        std::nullopt);
+#else
     realInputBuffer(node).add(
         input_nr,
         check_and_reduce(node->node, input_nr, std::move(t)),
         std::nullopt,
         std::nullopt,
         node->node);
+#endif
   }
 
   InputBuffer& realInputBuffer(NodeState* state) {