test(BA-7843): add scenario tests for the runtime variant adapter

[jopemachine]

Resolves BA-7843

Summary

• Fill api/adapters/runtime_variant/KNOWLEDGE.md with the scenarios the adapter guarantees, following the shape api/adapters/domain/KNOWLEDGE.md set in test(BA-7779): PoC of scenario-table tests through the adapter layer #14448, and transcribe it into tests/scenario/bai_scenario/manager/runtime_variant/ — 29 rows over five modules (creating, reading, searching, editing, retiring) — with seeds/runtime_variant/runtime_variant.py and components/runtime_variant.py.
• All eight calls are exercised, resolve_by_name and batch_load_by_ids included.

What the rows pin:

• The reads (by id, by name, by many ids) ask for authentication alone; a user granted nothing reads, and an unknown name resolves to not-found because no permission check follows the lookup.
• Create is behind the superadmin role; update and delete are behind the entity gate, which no role can reach since the variant is created in no scope. With rbac.enforcement_enabled=false a user granted nothing edits and deletes but still cannot create.
• Renaming to a taken name is refused with the raw unique-violation error, not the conflict error the create path maps it to.
• bulk_delete is a loop, not a transaction: an unknown id behind a known one is refused as not-found after the known one is gone.

Deferred, stated in the document: deleting a variant with presets attached — the preset's variant column has a foreign key in the alembic migration but not in the ORM, so the result differs between a migrated database and one built by schema oneshot.

tests/scenario/bai_scenario/components/system.py — the caller-only situation every system-concern table shares — travels in each of the eight sibling PRs byte for byte (#14535, #14536, #14538, #14542, #14543, #14544, #14545). Whichever merges first, the rest rebase clean.

Test plan

• python3 scripts/knowledge/check.py passes
• The rows passed in CI run 34558080395 of the unsplit branch (0 failing); report.md is generated from that run's artifacts with scripts/scenario-report.py --split
• test-scenario and verify-scenario-report pass on this PR

🤖 Generated with Claude Code

https://claude.ai/code/session_01Snp9H2jEVbB2s1C6ixUHGn