format on "[PyTorch] Avoid move-constructing a List in listConstruct"

List's move ctor is a little bit more expensive than you might expect, but we can easily avoid it.

Differential Revision: [D25542190](https://our.internmc.facebook.com/intern/diff/D25542190/)

**NOTE FOR REVIEWERS**: This PR has internal Facebook specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D25542190/)!

[ghstack-poisoned]

torch/csrc/jit/runtime/vararg_functions.cpp

@@ -211,15 +211,16 @@ void listConstruct(
   // Structuring the implementation this way allows NRVO to avoid
   // move-constructing vals on its way onto the stack. Moving a List
   // isn't free.
-  auto makeList = [](Stack& stack, const at::ListTypePtr& type, size_t num_inputs) {
-    c10::List<IValue> vals(type->getElementType());
-    vals.reserve(num_inputs);
-    for (size_t i = stack.size() - num_inputs; i < stack.size(); ++i) {
-      vals.emplace_back(std::move(stack[i]));
-    }
-    drop(stack, num_inputs);
-    return vals;
-  };
+  auto makeList =
+      [](Stack& stack, const at::ListTypePtr& type, size_t num_inputs) {
+        c10::List<IValue> vals(type->getElementType());
+        vals.reserve(num_inputs);
+        for (size_t i = stack.size() - num_inputs; i < stack.size(); ++i) {
+          vals.emplace_back(std::move(stack[i]));
+        }
+        drop(stack, num_inputs);
+        return vals;
+      };
   stack.push_back(makeList(stack, type, num_inputs));
 }