Flatten/Unflatten micro optimization in proxy_tensor.py (#121993)
Lowers compile time by 1s across all suites on average
Pull Request resolved: https://github.com/pytorch/pytorch/pull/121993
Approved by: https://github.com/anijain2305, https://github.com/jansel, https://github.com/zou3519
