Instantiates the dataset used during runtime of a Hive query. This means it's used in mappers and reducers,
and must use the Hadoop configuration to look up what dataset to instantiate. Should not be closed until the
dataset has been closed. Assumes the dataset name and namespace are settings in the configuration.
It may seem like this would not work if multiple datasets are used in a single query, but that is not the case.
It is not obvious, but dataset name and namespace are added as job properties in DatasetStorageHandler. This tells
Hive to add those properties to the Configuration object before passing it in to the methods of an InputFormat
or OutputFormat. So even if multiple datasets are used in the same query (a join query for example), dataset name
will not get clobbered.