`run_training_job` allows to pass `chunk_id` to switch data batch but our functions are not more generic than that

[610v4nn1]

          I changed back to using `chunk_id` to switch between the two datasets. I think it is a bit unintuitive for an example, when the `dataset_name` is part of the `config_space`. We explicitly introduced this `chunk_id` parameter in the `run_training_job` function to switch between chunks in the examples, so I think we should keep it consistent.

Originally posted by @lballes in #171 (comment)

We could generalize the function to be more flebible

[wistuba]

we already are more flexible. however, instead of passing it directly to run_training_job, the parameter needs to be an entry of config_space.