* feat(resilience): recovery capacity pillar — 6 new dimensions + 5 seeders (Phase 2 T2.2b)
Add the recovery-capacity pillar with 6 new dimensions:
- fiscalSpace: IMF GGR_G01_GDP_PT + GGXCNL_G01_GDP_PT + GGXWDG_NGDP_PT
- reserveAdequacy: World Bank FI.RES.TOTL.MO
- externalDebtCoverage: WB DT.DOD.DSTC.CD / FI.RES.TOTL.CD ratio
- importConcentration: UN Comtrade HHI (stub seeder)
- stateContinuity: derived from WGI + UCDP + displacement (no new fetch)
- fuelStockDays: IEA/EIA (stub seeder, Enrichment tier)
Each dimension has a scorer in _dimension-scorers.ts, registry entries in
_indicator-registry.ts, methodology doc subsections, and fixture data.
Seeders: fiscal-space (real, IMF WEO), reserve-adequacy (real, WB API),
external-debt (real, WB API), import-hhi (stub), fuel-stocks (stub).
Recovery domain weight is 0 until PR 4 (T2.3) ships the penalized weighted
mean across pillars. The domain appears in responses structurally but does
not affect the overall score.
Bootstrap: STANDALONE_KEYS + SEED_META + EMPTY_DATA_OK_KEYS + ON_DEMAND_KEYS
all updated in api/health.js. Source-failure mapping updated for
stateContinuity (WGI adapter). Widget labels and LOCKED_PREVIEW updated.
All 282 resilience tests pass, typecheck clean, methodology lint clean.
* fix(resilience): ISO3→ISO2 normalization in WB recovery seeders (#2987 P1)
Both seed-recovery-reserve-adequacy.mjs and seed-recovery-external-debt.mjs
used countryiso3code from the World Bank API response then immediately
rejected codes where length !== 2. WB returns ISO3 codes (USA, DEU, etc.),
so all real rows were silently dropped and the feed was always empty.
Fix: import scripts/shared/iso3-to-iso2.json and normalize before the
length check. Also removed from EMPTY_DATA_OK_KEYS in health.js since
empty results now indicate a real failure, not a structural absence.
* fix(resilience): remove unused import + no-op overrides (#2987 review)
* fix(test): update release-gate to expect 6 domains after recovery pillar
17e34dfca7