feat: split-style tracking issues and deferred creation for catalog-destined test cases

[ChristopherJHart]

Summary

• For catalog-destined test cases, project issues are now only created after a catalog PR exists, so the issue can reference it
• Catalog-destined test cases use the split-style tracking issue template (tracking_issue.j2) with catalog PR reference and task checklist, instead of the collapsed template (tac_issues_body.j2)
• Non-catalog test cases continue to use the collapsed-style template (unchanged behavior)
• Reorders the process_test_requirements loop: catalog PR creation → issue creation → project PR creation
• Labels are now applied to catalog PRs: adds quicksilver, removes script-already-created, keeps other labels (e.g., GenAI)
• Note: label application to catalog PRs was never implemented in any prior code path

Changes

Commit 1: Defer issue creation until catalog PR exists

• requires_issue_creation in test_cases_processor.py: checks metadata.catalog.destined and requires metadata.catalog_tracking.pr_number before allowing issue creation
• process_test_requirements in test_requirements.py: catalog PR creation moved before issue creation in the processing loop

Commit 2: Wire split-style tracking issue template

• create_tracking_issue_for_catalog_test_case() in test_requirements.py: renders tracking_issue.j2 from test case metadata, supporting both same-run and previous-run catalog PRs
• _extract_os_from_catalog_branch(): helper to extract OS from catalog branch name pattern feat/{os}/add-{stem}
• process_test_requirements loop branches on metadata.catalog.destined to choose split vs collapsed issue template

Commit 3: Apply labels to catalog PRs

• create_catalog_pr_for_test_case now accepts a labels parameter and calls set_labels_on_issue after PR creation (the GitHub PR creation API does not support labels inline)
• process_test_requirements passes test case labels through to catalog PR creation

Commit 4: Transform catalog PR labels

• Catalog PR labels are built from test case labels with script-already-created removed and quicksilver added
• Issue labels remain unchanged — they use whatever is defined on the test case

Tests

• 4 unit tests for catalog-destined issue creation gating
• 4 unit tests for _extract_os_from_catalog_branch
• 8 unit tests for create_tracking_issue_for_catalog_test_case
• 2 unit tests for catalog PR label application
• 2 unit tests for catalog PR label transformation (quicksilver added, script-already-created removed)
• 3 integration tests for end-to-end flow (same-run tracking issue, previous-run tracking issue, non-catalog collapsed template)

Test plan

• All 348 unit tests pass
• Manual validation against Google-GDC-XR-to-Nexus-Migration project in CI

---

🤖 AI Generation Metadata

• AI Generated: Yes
• AI Tool: claude-code
• AI Model: opus-4.6
• AI Contribution: ~85%
• AI Reason: defer issue creation + split-style tracking issues + catalog PR labels
• Human Oversight: Code reviewed and approved by chart2