Auto3d Add function to generate the shell scripts which can run separately.

[AHarouni]

Is your feature request related to a problem? Please describe.
Currently Auto3d generates a markdown telling you what command to run if you want to run a single network architecture
it would be nice to generate a bash script directly so user can directly run it

As a work around I go into the code and use the private function
cmd, devices_info = algo._create_cmd(train_param)
instead simply create the bash script to export cuda device=device_info then run the cmd as
(time $cmd )| tee log.txt

[myron]

thank you, are you trying to run a single algo?

you can simply call (from any generated algo folder)
python sripts/train.py run --config_file=configs/hyper_parameters.py
or for multigpu
torchrun sripts/train.py run --config_file=configs/hyper_parameters.py

the device_info is not needed (it's only to support running from notebooks).
the scripts folder already has train.py, eval.py, infer.py, if we add ".sh" one-liner aliases of these, it will a lot of files in the scripts folder

[wyli]

please see https://github.com/Project-MONAI/tutorials/blob/main/auto3dseg/docs/bundle.md