Improvements to segmentation tracking DB

[aamster]

• Currently, populating the DB happens outside of running the segmentation modules. It should instead be moved to the modules themselves so that it doesn't have to be done after the fact and is less error-prone.
• Add a new table called runs to track metadata about segmentation runs.
• The DB should know about what inputs were used. Log these in the DB, ie metric image, video, and all arguments used to produce outputs such as keep_fraction, etc.
• Add run_id and datetime. This will help keep track of multiple segmentation runs in the same DB.
• Add git commit to keep track of what code version was used
• Add more specificity to artifact artifact_type. Ie instead of background, it should be correlation_projection, max_projection, etc. The reason for this is to prevent duplicate entries in the DB. To further improve on reducing the possibility of duplicate entries, this table should have as primary key run_id, ophys_experiment_id, artifact_type.