Use the symint version of computeStorageNbytes within get_nbytes. (#101634)
Fixes [#ISSUE_NUMBER](https://github.com/pytorch/xla/pull/4998) according to [comment](https://github.com/pytorch/xla/pull/4998#issuecomment-1550232063).
This change is needed to make sure calling tensor.sizes() will error if the tensor has dynamic dimension in pytorch/xla.
Pull Request resolved: https://github.com/pytorch/pytorch/pull/101634
Approved by: https://github.com/ezyang
