doc: describe the --gres option, add link to StackOverflow

### GPUs

Jobs do not have access to the GPUs on compute nodes by default and users must explicitly request a specific number of GPUs for their jobs.

This can be done with the general `--gres` option as follows:

```bash

#SBATCH --gres=gpu:<number>

```

where the `<number>` specifies the number of GPUs _per node_.

Alternatively, different options for the allocation of GPUs are available, but they cause problems when running MPI jobs on multiple nodes (see [this question](https://stackoverflow.com/q/67091056) for details).

The following options are available:

| Option                        | Description |