[Slurm scheduler] Add better support for specifying resources in slurm

[aivanou]

🐛 Bug

According to aws/aws-parallelcluster#2198 PCluster has problems running jobs that have explicit memory requirements.

We need to modify our slurm scheduler to address this.

Module (check all that applies):

• torchx.spec
• torchx.component
• torchx.apps
• torchx.runtime
• torchx.cli
• [ x] torchx.schedulers
• torchx.pipelines
• torchx.aws
• torchx.examples
• other

To Reproduce

Steps to reproduce the behavior:

1. ssh to slurm cluster
2. create main.py that prints hello world
3. torchx run -s slurm --scheduler_args partition=compute,time=10 dist.ddp --script main.py

Expected behavior

Job successfully executed