Ingest Substack CSV

Workflow

Per weekly CSV:
- [ ] Step 1: Read header row; diff against .schema-expected.json
- [ ] Step 2: If load-bearing column missing, halt + emit schema-warning stub
- [ ] Step 3: Parse post rows (load-bearing: views, opens, open_rate, clicks, sent, post_date)
- [ ] Step 4: Parse subscribers aggregate (total, free/paid distribution, activity tiers) — NEVER row-by-row emails
- [ ] Step 5: Emit typed WeekExport object
- [ ] Step 6: On success, move CSV into corpus/stats/YYYY-WW.csv

Expected schema

Posts CSV columns (load-bearing in bold): id, slug, title, post_date, views, opens, open_rate, clicks, click_through_rate, sent, delivered, signups_within_1_day, subscriptions_within_1_day, unsubscribes_within_1_day, signups, subscribes, shares, estimated_value, engagement_rate, reaction_count, comment_count