Arm backend: Fix pte path in run.sh when using a model python file (#5979)

Summary:
This fix a problem when you use a file path as used model e.g. --model_name=<FILE_PATH.py>
Then the script didn't look for the generate pte in the same folers as the aot_arm_compiler.py put it.

Pull Request resolved: https://github.com/pytorch/executorch/pull/5979

Reviewed By: mergennachin

Differential Revision: D64047899

Pulled By: digantdesai

fbshipit-source-id: 770920a8fc83f613fdeeae6dad36ac2a3faaf00e