Create CordexML bench Data Index for PET #278

The CordexMLBench dataset creates a benchmark dataset for training and evaluating ML based climate downscalers for different regions of the world.

Tasks

• Create PET ArchiveIndex class for loading training data.
  *The data strcuture is somewhat unusual, so it might make sense for the class to be parameterised to only load 1 region and scenario at a time, as one can't merge different regions/scenarios into a single xarray dataset.
• Create a notebook to dekonstrate loading and visualising data
• Create a simple CNN based downscaler model for one region

Stretch Goals

• Create a PET training pipeline, using a training wrapper, for one or more of the models listed on the github page (depending on availability of training code)
• Create a inference wrapper for a model trained based on the models listed on the github page.

Further Info

• Cordex ML Bench on Zenodo
• CordexMLBench on Github