feat: Phase 1 — JA3/JA3S sniffer, Attacker model, profile worker

Add passive TLS fingerprinting via a sniffer container on the MACVLAN
interface, plus the Attacker table and periodic rebuild worker that
correlates per-IP profiles from Log + Bounty + CorrelationEngine.

- templates/sniffer/: Scapy sniffer with pure-Python TLS parser;
  emits tls_client_hello / tls_session RFC 5424 lines with ja3, ja3s,
  sni, alpn, raw_ciphers, raw_extensions; GREASE filtered per RFC 8701
- decnet/services/sniffer.py: service plugin (no ports, NET_RAW/NET_ADMIN)
- decnet/web/db/models.py: Attacker SQLModel table + AttackersResponse
- decnet/web/db/repository.py: 5 new abstract methods
- decnet/web/db/sqlite/repository.py: implement all 5 (upsert, pagination,
  sort by recent/active/traversals, bounty grouping)
- decnet/web/attacker_worker.py: 30s periodic rebuild via CorrelationEngine;
  extracts commands from log fields, merges fingerprint bounties
- decnet/web/api.py: wire attacker_profile_worker into lifespan
- decnet/web/ingester.py: extract JA3 bounty (fingerprint_type=ja3)
- development/DEVELOPMENT.md: full attacker intelligence collection roadmap
- pyproject.toml: scapy>=2.6.1 added to dev deps
- tests: test_sniffer_ja3.py (40+ vectors), test_attacker_worker.py,
  test_base_repo.py / test_web_api.py updated for new surface

templates/sniffer/decnet_logging.py

@@ -0,0 +1 @@
+# Placeholder — replaced by the deployer with the shared base template before docker build.