Geo Stage 4 (data layer): geo on feed responses + home-scope query filters

Foundation for "Closer to Home" (server-side, Codex-approved). No behavior change
yet — geo_scope defaults None, so the default/edge-cached feed is identical.

- queries.feed now returns each article's geo (breadth, confidence, and ISO-coded
  places) via a LEFT JOIN + places subquery. Article.from_row parses geo_places
  into [{country, state}]. Brief query doesn't select geo, so the Brief stays bare.
- queries.feed gains home-scope filters (home_country/home_state/geo_scope =
  near|country|world): STATE match only counts on high/medium geo confidence;
  untagged articles fall to 'world' so nothing is lost during backfill.

Next: API composition (home param + near/country/world sectioning with soft/blended
headers + a next_offset pagination model) and the Home picker UI. 360 tests green.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

goodnews/api.py

@@ -322,11 +322,26 @@ class Article(BaseModel):
     paywalled: bool = False
     tags: list[str] = []
     summary: str | None = None  # our own cached summary (present on the brief)
+    # Subject geography (present on feed rows; absent/empty on the brief). breadth is
+    # locality|regional|national|multinational|global|unknown; places are ISO codes.
+    geo_breadth: str | None = None
+    geo_confidence: str | None = None
+    geo_places: list[dict] = []   # e.g. [{"country": "US", "state": "NY"}, {"country": "GB", "state": None}]
 
     @classmethod
     def from_row(cls, row: dict) -> "Article":
         raw_tags = row.get("tags")
+        places = []
+        for tok in (row.get("geo_places") or "").split(","):
+            tok = tok.strip()
+            if not tok:
+                continue
+            cc, _, sc = tok.partition("-")
+            places.append({"country": cc, "state": sc or None})
         return cls(
+            geo_breadth=row.get("geo_breadth"),
+            geo_confidence=row.get("geo_confidence"),
+            geo_places=places,
             summary=row.get("summary"),
             id=row["id"],
             title=row["title"],