fix: complete netengine down command + fix registry handler tests

[aidankhogg]

Summary

netengine down was partially implemented — it only scanned for netengines_* containers, missing netengine_coredns and netengine_gateway which use a different prefix. Also lacked --dry-run, zone file cleanup, and pgmq queue purging.

Registry handler tests were broken — fixtures patched non-existent module-level get_supabase attributes and set h._pgmq instead of h.pgmq, causing live Postgres connection attempts in unit tests.

netengine down changes (netengine/cli/main.py)

• Add --dry-run flag: shows exactly what would be removed without touching anything
• Fix container scan to cover both netengine_ and netengines_ prefixes (coredns/gateway use the former, all others the latter)
• Use RuntimeState container IDs (dns_root_container_id, gateway_container_id, etc.) for precise targeted removal on top of the prefix scan
• Add zone file cleanup: removes NETENGINE_ZONE_DIR (Corefile + zones/) on teardown
• Add pgmq queue purge: calls pgmq.purge_queue() for all 6 queues and DLQs
• Convert down() to async via asyncio.run(_down()) to support awaitable cleanup steps

Test fixes (tests/test_registry_handlers.py)

• Pre-seed h._db = sb and h.pgmq = pgmq_mock directly on handler instances instead of patching module-level names that don't exist post-refactor
• Remove unused patch and AsyncMock imports

Test plan

• poetry run pytest — all 243 tests pass, 0 failures
• netengine down --dry-run lists containers/networks/zones/queues without removing anything
• netengine down with Docker running removes all netengine_* and netengines_* resources

🤖 Generated with Claude Code

---

Generated by Claude Code