#20875 fix updating of denormalized fields (_site, _location, _rack) for component models

[arthanson]

Fixes: #20875

Several cases were previously missed:

• netbox/dcim/models/devices.py - when components are instantiated from the device-template and the denormalized fields were never getting updated.
• netbox/dcim/models/modules.py - when creating a module and instantiating components
• netbox/dcim/signals.py - when changing the site for a location, when changing the site for a rack

Also checked the API and it looks like the API and this should also affect those. Bulk updates on the API do serializer.save() on each object which will cause the signal handler to run.

[jnovinger]

I think I agree with @jeremystretch's suggestion about populating fields in instantiate(). I see two distinct problems here that may not need the same solution:

1. Device/Module creation

When components are instantiated via bulk_create(), the denormalized fields are NULL. Jeremy's suggestion handles this populate _site, _location, and _rack in each template's instantiate() method so the fields are set correctly before bulk_create() is called. No post-processing UPDATE needed.

2. Location/Rack changes (signal handlers)

The helper function creates an O(m * n) query pattern here. When a location's site changes with 100 devices, this executes 1,000 UPDATE queries (100 devices × 10 component models). The signal handlers could instead bulk-update per component model (see suggestion in other comment for how I'd approach that).

Given that needed change of behavior, I don't think the helper function provides enough value to justify abstracting these two distinct patterns. Case 1 can be handled in instantiate(), and Case 2 needs efficient bulk updates.

Although, perhaps I'm missing something here. Let me know.