Consider making export time stamps more granular

[landreev]

In the current implementation a dataset entity has a single lastexporttime time stamp. Meaning, it's all-or-nothing, a dataset is considered either exported, or not. Seeing how we now support a dozen different metadata formats, this should probably be made more granular. It happens all the time that it's just one specific format that fails to export. So, for the purposes of, for example, serving OAI sets, it would make more sense to be able to select all the datasets for which the OAI_DC, OAI_DDI etc. formats have been successfully exported - but not necessarily ALL of them.

Similarly, under the current scheme whenever we make some change that affects just one metadata format, we tell the installations that a full reexport of all the formats is required. Which feels like quite a waste.

[pdurbin]

Notes from tech hours:

Store timestamps per export format (DDI, Croissant, etc.)

Consider switching to aux files subsystem. Aux files are registered in the database. Determine if cache exists. More database calls, but that's ok.

We think it would be an 80 due to complexity, etc. but we are timeboxing this as a smaller, time-boxed size (10) for investigation as a spike.