Support multi-session quality_control by tracking metric or evaluation provenance

[dbirman]

Users have requested that we find a way to track multi-session quality control in the main schema, rather than some kind of extension.

Short explanation for why this is needed:

• A user has ten assets that all do their own raw/processing qc and pass and then they need to compare them somehow, e.g. matching ophys fov
• They run a multi-session capsule that generates evaluations/metrics to check whether the fov is the same. This is a blocking step because the annotation is manual
• Someone annotates the metrics and marks that one asset has a different fov, it needs to be dropped for further multi-session analysis
• A second multi-session is now run that needs to pull only the assets with the same fov. To match metrics -> assets the metrics need to track which asset they were generated from

[dbirman]

Notes from discussion on 10/3:

• There's no easy ability to generate isolated data assets the way I had hoped for, so dumping the quality_control.json isn't that simple, people are going to have to use the data-access-api to do this, with some limited permissions
• We need to stratify QC status based on local vs global scope and include whether evaluations have multiple upstream dependencies
• We need to mark evaluations (or metrics?) as having external dependencies, and possibly list their dependency structure using the "name" fields