Revenue breakdown by operating segment, geography, and product line.

Data source priority:

1. PostgreSQL segment_revenue table (cached)
2. On-disk *_xbrl.json file (extracted on demand, then cached to PG)

Response shape::

{
  "ticker": "AAPL",
  "operating": [{"name": "iPhone", "years": [{"year": 2023, "revenue": 200583000000}, ...]}, ...],
  "geographic": [...],
  "product": [...],
  "source": "pg_cache" | "pg_cache_refreshing" | "xbrl_live",
  # pg_cache_refreshing: these rows came from an OLDER extractor and a
  # background re-extraction was just started (v229). The rows are real;
  # a later request carries the improved ones.
  "has_data": true
}

Each segment has a years array sorted newest-first so chart rendering is straightforward.