Deterministic, no-LLM market metrics from Objective data for a location
(district and/or obj_ids): absorption_rate, months_of_supply, sell_through_pct,
unit/area_velocity, liquidity_index per room-type, overstock_index,
demand_concentration (HHI), price_sensitivity (reuses
analytics_queries._elasticity_coef). Frozen MarketMetrics dataclass + as_dict,
sample-size confidence (§15 spirit).
Graceful on thin data (the #1 coverage risk): no data → metric None (never
0/crash), confidence='low'. Filters by district/obj_ids to sidestep the sparse
domrf↔objective mapping. Read-only SELECTs, 55 unit/mock tests. Consumed by
forecast (#952) + relevance (#949 PR B). No migration, no endpoint.
