Send pipeline names in benchmark arguments

[sarahmish]

PR #465 introduced a new feature to the benchmark which allows it to be run from the command line with arguments. The argument pipelines accepts a list of pipeline names but we would like it to also allow a dictionary of pipelines like in the original proposal.

A possible issue that might arise when keeping the current state is that loading the hyperparameters per dataset will be impacted. To prevent this issue, we can have two solutions:

• Parse arguments as dictionaries to have a clear pipeline name.
• Augment the pipeline names in a "fixed" fashion, e.g. pipeline_name = 'tadgan' if pipeline == 'tadgan_gpu'.
• Have a new parameter in the JSON file to give an alias name for the pipeline.