Fix exit status handling for all GPU ML Jobs
At certain scenarios when identifying container exit statuses, it may lead to false truths in cases where the invoked GPU ML job(ml_job.py) encounters an error, and the ML job preparer(prepare_ml.py) does not handle the exit code, making the container exit status appear as successful (exit code 0) even if the invoked job failed.
Edited by Avimanyu Bandyopadhyay