Support for a job creating multiple datasets/multiple of an artifact type

[njbrake]

Currently, our inference job (which includes the annotation job) and backend code assumes that only a single dataset will be created. We should refactor our code such that we can support multiple datasets being stored. This point came up during PR-815

See comment https://github.com/mozilla-ai/lumigator/pull/815/files#r1948580576