Canonicalize + register Panama food_acquired — needs data_scheme + sample scaffolding (Phase 3 of #218)

[ligon]

Goal

Modernize Panama food_acquired (canonical + registered) → derived tables surface → retire dead legacy script.

Why

food_acquired.py exists (country + 1997/2003/2008 waves) in legacy (j,t,i,u) shape; unregistered → modern API empty. Legacy pqe dead (KeyError('m')).

Readiness: LARGER — scaffolding first

Panama has NO data_scheme.yml and NO sample table. Before food_acquired can be registered + v-joined:

1. Author a data_scheme.yml.
2. Build a sample table (cluster identity from upm).
   Then canonicalize food_acquired + register + acceptance bar (canonical + sanity + source reconciliation + Feature named) + retire legacy (same PR).

Parent: #218. Design: slurm_logs/DESIGN_218_food_legacy_retirement.org (rev-2). Adversarial review: workflow wuiofz6p0 (11 confirmed findings).

[ligon]

Verified fixed (2026-07-12) — closing.

Country('Panama') now builds all four food tables clean:

table	sane
food_acquired	✅
food_expenditures	✅
food_prices	✅
food_quantities	✅

sample() present (18353 rows), and Panama is registered in catalog.countries() so Feature() assembles it.

Verified against a live build (is_this_feature_sane clean on each). Reopen if the coverage matrix disagrees on a clean-cache run.

⚠️ Note for future reconciliation: a sane cell is not proof an issue is fixed — the matrix is a cold build and does not check currency labels (cf. #588, #589, which look green and are genuinely open). These four were verified by building the tables directly, not by reading a tier.