[Documentation] Call shell scripts with extra resources inside the submitted Python function

For SLURM each call of srun creates a new job step, so there is no need to assign any resources to the specific Python function, just assign the resources to the external shell script directly.

For flux there is the option to use nested executors:
https://executorlib.readthedocs.io/en/latest/3-hpc-job.html#nested-executors

Still this option can also be assigned during the submission - just like all the other options:

from executorlib import FluxJobExecutor

def get_available_gpus(lst):
    import socket
    from tensorflow.python.client import device_lib
    local_device_protos = device_lib.list_local_devices()
    return [
        (x.name, x.physical_device_desc, socket.gethostname())
        for x in local_device_protos if x.device_type == "GPU"
    ] + lst

with FluxJobExecutor() as exe:
    fs = []
    for i in range(1, 4):
        fs = exe.submit(
            get_available_gpus,
            lst=fs,
            resource_dict={"cores": 1, "gpus_per_core": 1, "flux_executor_nesting": True},
        )
    print(fs.result())

Start with the linked nested-executors documentation and the repository’s existing HPC job documentation; no specific file is named in the issue. Document how SLURM shell scripts and Flux nested executors receive extra resources at submission time, using the shown Python example as the reference for what done looks like.