Audit Model Target mock-only tests

[adamtheturtle]

Follow-up to #3191.

Some tests/mock_vws/test_model_target_web_api.py::TestMockErrors tests are intentionally mock-only today because they were added during a CI coverage fix. In the long run, externally observable Model Target Web API behavior should be verified fake coverage against real Vuforia wherever practical, and mock-only tests should remain only for internal mock mechanics or cases that are explicitly documented as impractical to verify against real Vuforia.

Acceptance criteria:

• Audit every Model Target mock-only test and classify it as API-observable behavior or internal mock mechanics.
• Convert or replace API-observable mock-only tests with verified fake tests where practical.
• Link any converted cases to the relevant behavior-specific issues (Improve Model Target Web API authentication fidelity #3192, Match Model Target dataset creation validation #3193, Improve Model Target dataset status fidelity #3194, Make Model Target dataset downloads more realistic #3195, Add verified fake coverage for advanced Model Target workflows #3196).
• Document the rationale for any remaining mock-only Model Target tests, either in the test name/comment or in docs/source/differences-to-vws.rst.
• The final suite has no unnecessarily mock-only Model Target tests.