fix(router.py): skip setting model_group response headers for now

current implementation increases redis cache calls by 3x

tests/local_testing/test_completion.py

@@ -3069,6 +3069,7 @@ def test_completion_azure():
             api_key="os.environ/AZURE_API_KEY",
         )
         print(f"response: {response}")
+        print(f"response hidden params: {response._hidden_params}")
         ## Test azure flag for backwards-compat
         # response = completion(
         #     model="chatgpt-v-2",