Add pipeline registry

[ellwise]

To make datasets discoverable - could wrap kedro's Pipeline class in another class that includes a class method and "global" variable for listing all pipelines that have been created. Could also introduce some way of telling people which other pipeline should be run if a dataset is missing (i.e. catch the exception).