Estimator ADDITIVELY derives expected_sold from asking-median × per-rooms
asking_to_sold_ratio (table from #648 S2 / migration 080). Headline asking
median, ranges, confidence and actual_deals are UNCHANGED. Graceful: missing/
empty ratio table → expected_sold fields None, estimate still returns 200
(ratio lookup rolls back the shared session on error so the subsequent INSERT
isn't poisoned — mirrors sibling cache helpers).
- estimator: _get_asking_sold_ratio(db, rooms) cached lookup (per-rooms →
global -1 fallback → None, try/except+rollback), applied after repair-coef.
- schema: 6 optional expected_sold_* / ratio fields on AggregatedEstimate.
- api: rehydrate the new columns in GET /estimate/{id} + /pdf reopen SELECTs
and both constructors (the documented reopen-path bug).
- migration 081: ADD COLUMN IF NOT EXISTS for the persisted subset.
- tests: 10 new (lookup/apply/graceful) + fix pre-existing test_estimator_
source_quota 2→3 tuple unpack of _fetch_analogs (was RED on main).
