GatherMetadataJob -- populate subject_details.reward_consume_*

[dbirman]

When these fields are empty, aggregate the total reward consumed from the individual stimulus_epochs and put that data here. Probably default to converting to ml?