feat: wire up unit status with presence

[SimonRichardson]

This is the last of the PRs to wire up unit status (agent and workload) with the presence data. This is quite easy now that we have the presence information in the database, the tricky part is ensuring that we correctly handle the status for things like allocating. If the unit is allocating, it won't yet have recorded the presence, in that case, we have to fake the display status when calling juju status. This won't impact the underlying data, just the display nature.

---

This pull request includes several changes to improve the handling of unit and application statuses in the codebase. The most important changes involve adding new methods to retrieve both unit and agent display statuses, modifying existing status retrieval methods to handle cases where status information might be missing, and updating the corresponding tests and mock implementations.

New Methods and Enhancements:

• apiserver/facades/client/client/service.go: Added GetUnitAndAgentDisplayStatus method to retrieve both unit and agent display statuses.
• apiserver/facades/client/client/status.go: Updated processUnitAndAgentStatus to use the new GetUnitAndAgentDisplayStatus method.

Status Handling Improvements:

• domain/application/service/application.go: Modified GetApplicationStatus and GetApplicationDisplayStatus to handle cases where applicationStatus is nil and to use the new UnitStatusInfo structure. [1] [2]
• domain/application/service/migration.go: Updated GetApplicationStatus and GetUnitAgentStatus to handle missing status information and to use the UnitStatusInfo structure. [1] [2]

Test Updates:

• domain/application/service/application_test.go: Updated tests to use UnitStatusInfo instead of StatusInfo. [1] [2] [3]
• domain/application/service/migration_test.go: Updated tests to use UnitStatusInfo and handle new status cases. [1] [2]

Mock Implementations:

• domain/application/service/package_mock_test.go: Updated mock methods to use UnitStatusInfo instead of StatusInfo. [1] [2] [3] [4] [5]

These changes collectively enhance the robustness of status handling in the application, ensuring that missing status information is appropriately managed and that the new UnitStatusInfo structure is consistently used across the codebase.

QA steps

It should switch between the statuses correctly.

$ juju bootstrap lxd test
$ watch juju status -m controller

Deploying an application to a model, should correctly display allocating and not agent lost.

$ juju add-model default
$ watch juju status

In another terminal (as it can be fast!):

$ juju deploy ubuntu

Bonus points

Whilst still calling watch juju status:

$ juju ssh 0
$ sudo systemctl stop jujud-machine-0.service

Status should now switch correctly to lost for the agent and unknown for the application and workload:

$ sudo systemctl start jujud-machine-0.service

It should come back - you can flip-flop between the two states.

Links

Jira card: JUJU-7652