feat(demo): repin serving demo store to the kitchen-appliance legend (B3)

[brownjuly2003-code]

Summary

• Repins the four serving demo tables (products_current, orders_v2, users_enriched, sessions_aggregated) from the grocery/electronics-era placeholder values to the kitchen-appliance legend, per docs/generator-spec.md §9–10: 10 products (one per SKU category, one out-of-stock bestseller for the cross-channel oversell story), 8 orders in RUB (5 marketplace 1.5–3k + 1 D2C ~4k + 2 wholesale ~48k/~76k — bimodal per spec), 5 users (2 dealer + 3 retail), 6 sessions remapped to the retail-only D2C telemetry story. Mirrored identically across duckdb_backend.py and clickhouse_backend.py.
• Propagates RUB as the revenue/avg_order_value metric unit end to end: catalog.py (the live source), regenerated config/contracts/metric.*.yaml via scripts/generate_contracts.py (fixed the drift-check generator, not hand-edited), contracts/entities/order.yaml currency description, and search-index snippet text.
• Updates the handful of tests/docs that asserted the old literal names, prices, counts, and unit: "USD" strings — found by running the full suite and fixing only what the demo-store change actually broke (structural stub-based tests using "USD" as an arbitrary placeholder were left untouched).

Scope notes

• Left the ingestion pipeline's Currency enum (USD/EUR/GBP) and config/contracts/order.v1.yaml/order.v2.yaml untouched — that governs synthetic CDC/event validation, decoupled from the static demo seed, and out of B3's stated scope (serving demo store + NL demo answers + README examples).
• Left docs/dv2-multi-branch/demo_evidence.md, perf reports, and notebook output cells untouched — those are live-stand-generated evidence and belong to B4, not something to hand-edit.

Test plan

• pytest tests/unit — 1587 passed (2 pre-existing environmental flakes unrelated to this change: a Windows thread-timing test and a local helm-binary blip, both reproduced independently of this diff)
• pytest tests/integration tests/property -m "not requires_docker and not live" — 217 passed, 50 skipped, 5 deselected
• ruff check + ruff format --check on all touched files
• mypy src/ — clean

🤖 Generated with Claude Code

[github-actions]

DORA Metrics

• Window: last 30 days
• Branch: main
• Deployment frequency: 116 total / 27.07 per week
• Lead time for changes: avg 0.44h / median 0.0h
• Change failure rate: 81.03% (94/116)
• MTTR: 0.29h across 2 incident(s)