ENH: Add support to specify input and output data names

[fepegar]

When downloading or mounting a data asset on AML, it can be given a "name" that can be used to retrieve the target folder dynamically inside the submitted job. This PR adds support to specify names, alternatively to the default INPUT_0 etc.

Before:

dataset_config = DatasetConfig(
    name=path,
    version=version,
    use_mounting=mount,
)

After:

dataset_config = DatasetConfig(
    name=path,
    version=version,
    use_mounting=mount,
    data_name="eval_data_dir",
)

The command may now include an argument ${{inputs.eval_data_dir}} which will be interpolated to the local folder where the data lives during the run.

[ant0nsc]

Overall this looks great and is a nice change in particular for SDK v2 jobs.
It would be nice if you could add documentation here https://hi-ml.readthedocs.io/en/latest/datasets.html that shows this flag in action.

[fepegar]

Thanks @samb-t and @ant0nsc for reviewing!