MPT-21922 finalize helpdesk queues/cases on E2E teardown (+ _process_value refactor)

[jentyk]

This branch carries two changes under MPT-21922.

---

1. MPT-21922 — E2E helpdesk teardown cleanup (the ticket)

Bug: "[E2E] Queue and Case Status update upon test completions — when test completed, update queue status to disabled and case status to completed." (reported by John McDonnell)

E2E helpdesk fixtures left resources in non-terminal states after a test:

• created_case / async_created_case had no teardown at all, and CasesService has no delete — so every run left cases lingering open.
• created_queue tore down via delete, but case_data attaches a case to that queue; deleting a queue with an active case fails (the "TEARDOWN - Unable to delete" path), leaving the queue active too.

Fix (tests/e2e/helper.py, tests/e2e/helpdesk/conftest.py):

• Add create_fixture_resource_and_finalize + async_create_fixture_resource_and_finalize, mirroring the existing delete-teardown context managers but invoking a terminal-state transition best-effort (errors logged, never raised — same convention as the delete helpers).
• Wire helpdesk fixtures: queues are disabled and cases are completed on teardown. Fixture ordering means the case is completed before its queue is disabled. Forms and parameters keep their existing delete teardown.

Note: these are E2E tests against the live MPT API, so they require credentials to execute and were not run as part of this change. Verified via make check and pytest --collect-only (32 tests collect cleanly). Teardown is best-effort, so a case that can't reach Completed from its current state logs and moves on rather than failing the test.

---

2. _process_value cognitive-complexity refactor

Refactors BaseModel._process_value in mpt_api_client/models/model.py to reduce its Cognitive Complexity from 18 to ~4 (allowed: 15), flagged by SonarQube.

The dict- and list-conversion branches are extracted into two flat module-level helpers (_build_model_from_dict, _resolve_list_model_class), removing the nesting that drove the score. Behaviour is unchanged; the redundant trailing isinstance(value, BaseModel) check folds into the final return. Helpers are kept module-level (matching to_snake_case/to_camel_case) to avoid tripping WPS214's method-count limit.

---

Verification

make check passes clean for both changes (ruff format, ruff check, flake8/WPS, mypy, uv lock --check); 118 model unit tests pass; helpdesk E2E suite collects cleanly.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
100.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud