Consolidate external reference fields for externally-sourced model data

[jmurty]

In GLAMkit we have a number of models that can be imported from external data sources: images, events and occurrences, collection models.

Currently all these models share an external_ref character field that, by convention, is used to store an external reference that we can use in API operations or data import scripts to uniquely identify and cross-reference records imported from external systems.

However, there are two problems with this approach:

• the external_ref field itself is defined in multiple places, and can have different characteristics, such as maximum size values (Image.external_ref is larger since #251). These inconsistencies are likely to confuse and/or stymie us in the future
• the single external_ref character field is not always enough information to disambiguate and cross-check records between GLAMkit and external systems. In some cases we bake external system modification times into this field, where such data should instead be in extra DB fields specific for that purpose.

I propose that we consolidate all the externally-referrable models to be consistent and DRY by:

• add a base abstract model class mixin for all externally-referrable models
• include the external_ref field in that base model, and update all other referrable models to use the base model:
• add whichever other fields we need to properly handle situations where we need to remember modification times from external systems, perhaps by adding nullable external_dt_modified and external_dt_created fields

As part of the consolidation, we could also take the chance to:

• rename the external_ref field to something clearer like external_reference or external_identifier
• increase or adjust the size of the field to better match real requirements, and make the size consistent across all models.

The main drawbacks with all these changes is it would mean:

• an avalanche of DB migrations
• we could no longer have specific help text for each external_ref (or similar) field, since the help text will be defined in the base class and will therefore need a generic description
• with centralised external reference fields is that they would apply to Occurrence models in events, which should ideally be as lightweight as possible. Extra date time fields and a large external reference field would work against that

[markfinger]

In the case of the AGSA sync for image files/records, the external reference's continuing to get murkier, as it now relies on four values:

• FS path
• FS mtime
• EMU id
• EMU mtime

The EMU records contain data such as title and rights, so we need to sync these sporadically.