fix(dv2): rename record_source example x5__* -> mp__* (B2)

[brownjuly2003-code]

Summary

• The customer-MDM admission logic (audit_28_06_26 chore(deps,sdk-ts): bump vitest from 3.2.4 to 4.1.7 in /sdk-ts #12) proves it is source-agnostic by citing a third record_source convention alongside 1c__/pg_ops__. That example carried the name of the Kaggle seed dataset (X5 Retail Hero); under the legend (domain.md §5.3-5.4) it is the consolidated marketplace feed, mp__.
• Renamed everywhere it appears: the 5 ClickHouse bv_customer_mdm__* view header comments, the PostgreSQL port (03_business_vault.sql), the PostgreSQL governance live-verify seed literal (postgres/governance/verify_live.sh — the one place x5__ was an actual value, not just prose), and both admission-test docstrings (test_dv2_business_vault_ddl.py / test_dv2_postgres_ddl.py).
• Added a ratchet assertion (x5__ absent / mp__ present) to each admission test, checked on raw file text since the example lives in a header comment the tests otherwise strip for the parse-logic checks.
• docs/domain.md §5.3/§5.4 updated: the record_source prefix table drops the (currently x5__) note, and the planned-rename row is marked Done.

x5__ was never a real record_source value produced by the generator/seeds (those emit 1c__/pg_ops__/wb__) — it only existed as documentation vocabulary for this one audit story, so this is a pure rename with no data/behavior change.

Test plan

• pytest tests/unit/test_dv2_business_vault_ddl.py tests/unit/test_dv2_postgres_ddl.py -q — 28 passed
• pytest tests/unit -q — 1589 passed
• ruff check / ruff format --check on all touched files — clean

[github-actions]

DORA Metrics

• Window: last 30 days
• Branch: main
• Deployment frequency: 115 total / 26.83 per week
• Lead time for changes: avg 0.44h / median 0.0h
• Change failure rate: 73.91% (85/115)
• MTTR: 0.29h across 2 incident(s)