Add cached_method decorator and refactor Flight views/clicks caching

[Copilot]

• Understand the existing code structure and PR Remove aggregation from critical path - use local cache #1150
• Add cached_method decorator to adserver/utils.py
• Refactor Flight.views_today() and Flight.clicks_today() to use @cached_method
• Add tests for cached_method in test_utils.py
• Add DB query count tests for views_today/clicks_today in test_models.py
• Run linting (pre-commit run --all-files) - all hooks pass
• Run full test suite - all tests pass (1 pre-existing unrelated failure in test_ad_bulk_create_view)
• CodeQL security check - no alerts
• Reduce FakeModel duplication in CachedMethodTest via _make_obj factory helper

---

🔒 GitHub Advanced Security automatically protects Copilot coding agent pull requests. You can protect all pull requests by enabling Advanced Security for your repositories. Learn more about Advanced Security.

[ericholscher]

Looked over this and it seems generally ready.

@copilot reduce the fake model duplication in the tests if it isn't too clunky.

[ericholscher]

@copilot research other places in the code where a similar approach would work well, and open a new PR adding the decorator there.

[Copilot]

Looked over this and it seems generally ready.

@copilot reduce the fake model duplication in the tests if it isn't too clunky.

Done in 3e28d82. Each test now calls self._make_obj(**attrs) which defines FakeModel and the decorated method once, then sets pk and any annotated attributes on the returned instance.

@copilot research other places in the code where a similar approach would work well, and open a new PR adding the decorator there.

I'm not able to open new PRs from this environment. Looking at the codebase, the most promising candidate is Campaign.total_value() (models.py ~line 740) which already has a manual hasattr(self, "campaign_total_value") guard but no local-cache layer — it would benefit directly from @cached_method("campaign_total_value"). The Advertisement.total_views() / total_clicks() methods (models.py ~line 2307-2317) also do bare aggregations with no caching at all. Happy to add those in a follow-up if you'd like to open a tracking issue or point me at a branch.

[davidfischer]

I gave a cursory look and this looks good. I'm not planning to merge this today for the release. Specifically, what I'm most concerned about, is the admin dashboard performance where we're fetching 100s of flights at a time and we had an in-built optimization there that this removes. There's probably a quick fix for that though.