[DOC-27] Add an multi-node example based on Accelerate

[lebrice]

Example of how to use accelerate with SLURM, based on https://github.com/huggingface/accelerate/blob/main/examples/by_feature/checkpointing.py

Compared to that example, here are the differences (benefits):

• This example demonstrates how to implement code checkpointing: jobs are executed with the code as it was at the time of job submission, even if the code is updated later.
  • Clones the repository at the commit at the time of job submission in $SLURM_TMPDIR
  • Uses uv run in that directory to recreate the virtualenv in $SLURM_TMPDIR on each node in a multi-node job.
  • This is done by submitting the job with the safe_sbatch script instead of sbatch.
    Compared to sbatch, safe_sbatch will prevent submitting jobs if there are uncommitted
    changes in the code repository.

[obilaniu]

Même principe ici - est-ce que les choses dans uv.lock sont d'une réelle importance à l'exemple?

[btravouillon]

As per https://slurm.schedmd.com/faq.html#acronym, use "Slurm", not "SLURM".

Suggested change

	2. On SLURM clusters where you do not have internet access on compute nodes, you need to first create the virtual environment:
	2. On Slurm clusters where you do not have internet access on compute nodes, you need to first create the virtual environment: