slurm: support clusters without sacct

[d4l3k]

Many bare-bones slurm clusters aren't configured with accounting enabled (sacct) as it requires an external SQL database. This makes torchx slurm scheduler list/describe/status methods fallback to squeue which will show currently running jobs.

It also shows the job name in torchx list as slurm job_ids are just numbers

Test plan:

pytest torchx/schedulers/test/slurm_scheduler_test.py

(pytorch-3.12) ubuntu@slurm-head-node-0:~/tristanr/runner$ torchx status slurm://torchx/163
Slurm accounting storage is disabled
AppStatus:
    State: RUNNING
    Num Restarts: -1
    Roles: 
 *train[0]:RUNNING
    Msg: RUNNING
    Structured Error Msg: <NONE>
    UI URL: None
    
(pytorch-3.12) ubuntu@slurm-head-node-0:~/tristanr/runner$ torchx list -s slurm
Slurm accounting storage is disabled
APP HANDLE          APP NAME    APP STATUS
------------------  ----------  ------------
slurm://torchx/163  train-0     RUNNING

[facebook-github-bot]

@kiukchung has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.