test: improve unit test coverage toward 80% target

[JohnRDOrazio]

Problem

OntoKit API has 136 tests at 40% code coverage. Major areas are untested:

• collab/ module: 0%
• Core services (project, PR, ontology, git): 10-27%
• Route handlers: 23-54%
• Worker tasks: 17%

Plan

Add ~200 tests across 22 new files + set up integration test infrastructure in CI.

Phase 1: Foundations

• Add PostgreSQL (pgvector/pgvector:pg17) and Redis service containers to CI workflow
• Add shared test fixtures (mock_arq_pool, bare_git_repo, authed_client)

Phase 2: Pure Unit Tests (Tier 1 — highest ROI)

• collab/presence.py: 0% → 95%
• collab/transform.py: 0% → 95%
• collab/protocol.py: 0% → 90%
• core/encryption.py: 35% → 95%
• services/notification_service.py: 34% → 80%
• services/user_service.py: 17% → 75%
• services/github_service.py: 36% → 75%

Phase 3: Service Tests (Tier 2)

• services/project_service.py: 15% → 55%
• services/ontology_index.py: 13% → 55%
• services/ontology.py: 27% → 55%
• services/remote_sync_service.py: 21% → 65%
• services/join_request_service.py: 18% → 60%

Phase 4: Route Tests (Tier 3)

• api/routes/auth.py: 24% → 70%
• api/routes/lint.py: 24% → 60%
• api/routes/notifications.py: 73% → 92%
• api/routes/quality.py: 38% → 65%
• api/routes/user_settings.py: 33% → 65%

Phase 5: Git + Worker Tests (Tier 4)

• git/bare_repository.py: 23% → 55%
• worker.py: 17% → 45%
• Integration test: git operations with real pygit2 repos