feat: add AU and US geometry + region modules

[sperrot]

Summary

Contributes two new country modules for Australia and the United States, following the ATPolygonIndex interface already established in scripts/_lib/at/geometry.py (same resolve / union_all / n_pdo_polygons contract, same (geometry, geom_source, stats) return type).

AU — scripts/_lib/au/

• geometry.py — AUPolygonIndex backed by the Wine Australia GI Register (official, publicly available SHP/KML). Slug-based name matching handles minor spelling variants across the Zone / Region / Subregion hierarchy (~65 GIs). geom_source = "wine-australia-gi". No Bétard-2022 fallback (EU-only); unresolved GIs return "stub-no-geometry".
  Expected path: raw/au/wine_australia/gi.shp
  Source: https://www.wineaustralia.com/labelling/register-of-protected-gis

• region.py — derive_state() — returns a 2-letter AU state code (NSW, VIC, QLD, SA, WA, TAS, ACT, NT). Multi-state GIs return the first listed state (Wine Australia convention).

US — scripts/_lib/us/

• geometry.py — USPolygonIndex backed by the UCDavis Library AVA corpus (CC0). Keyed by ava_id (e.g. alexander_valley_19841123). Covers all 245 TTB-registered AVAs. geom_source = "ucavis-ava".
  Expected path: raw/us/ucavis/avas.geojson
  Source: https://github.com/UCDavisLibrary/ava (CC0)

• region.py — derive_state() — returns a 2-letter US state code. Multi-state AVAs (e.g. state = "OR|WA") return the first token (TTB petition convention).

Scope

Geometry and region derivation only — no wiki pipeline content, no eAmbrosia integration (AU/US are outside the EU GI register). These modules slot into stage 04 in the same way as existing EU country modules.

Test plan

• AUPolygonIndex("raw/au/wine_australia/gi.shp").resolve("Barossa Valley") returns a polygon with geom_source == "wine-australia-gi"
• AUPolygonIndex.resolve("unknown-gi") returns (None, "stub-no-geometry", {"matched": 0, "unmatched": 0})
• USPolygonIndex("raw/us/ucavis/avas.geojson").resolve("alexander_valley_19841123") returns a polygon
• derive_state({"state": "OR|WA"}) returns "OR" (US); derive_state({"state": "NSW|VIC"}) returns "NSW" (AU)
• find_state_in_text("appellation in South Australia") returns "SA" — longest-match prevents false positive on NSW

🤖 Generated with Claude Code

[bdevloed]

Thanks for this contribution!

I'll build on top of this for to add the terroir extraction, allowed grapes etc. (if available - US regulation is a bit more relaxed if I remember correctly).

Timeline: coming weeks.