[BUG] NIP-56 report ingestion: dead rows, WoT-rebuild blocking, target truncation, no transaction #785

Describe the bug

1. A kind-1984 event with no valid p/e target still inserts a reports row with nothing to act on (src/utils/nip56.ts, src/handlers/event-strategies/report-event-strategy.ts).
2. The first report accepted after a worker restart blocks on WotGraphService's full graph rebuild via getDistance() -> ensureBuilt(), a one-time latency spike on the event-accept path.
3. extractReportTargets only reads the first p and first e tag -- a report naming many pubkeys silently records just one, with no cap or log line.
4. The event insert and target-row insert(s) in ReportEventStrategy aren't wrapped in a transaction -- a partial failure leaves some rows missing, and the existing per-event duplicate guard means a re-sent report never retries, so the loss is silent beyond a log line.

Expected behavior

• No-target reports are skipped, not stored.
• Report ingestion doesn't block on a cold-start WoT graph rebuild.
• All p/e targets on a report event are recorded (or the single-target limit is documented).
• The event and its report row(s) are inserted atomically.