All models: Provide a mechanism for reprocessing data

[colmsnowplow]

If there's some issue in the atomic data for a period of time (eg failed events) that gets fixed subsequently to a model running over the time period, reprocessing involves manual intervention and is more sensitive than we'd like it to be - mostly due to manifests.

We could provide some mechanism to reprocess data for any given window of time, which hopefully would make UX much simpler in such cases.