Adding default xml dir. Making batching size-based rather than number-based. Adding submit action. Making order of runs.csv deterministic, removing 'sample_' prefix. Fixing experiment and dataset creation. Adding non-empty test data. Updating Conda environment.