Include input datasets as a field which can be returned from a query

[JoanneBogart]

Input datasets may be found by joining the dataset table with the dependencies table (just as is now done to return execution properties). However since more than one input may be associated with a dataset, some thought should be given to the form of the returned results.

[JoanneBogart]

There is code (untested and appears to be wrong) to create the entries in the dependency table, but that table currently associates a dataset with its dependency datasets. If an execution has several outputs, each would have similar entries in the dependency table, which seems wasteful. The dependencies are really associated with the execution. However currently the execution_id argument to register_dataset is optional and there might be legacy datasets where no true execution entry can be made. In this case (no execution id supplied) register_dataset could make a minimal entry in the execution table.

[JoanneBogart]

With the merge of u/jrbogart/dependencies the dependencies are now properly stored in the db.