v1.0.1

First post-v1.0.0 patch release. Adds operator-facing tuning for
SNMP polling and ships UPGRADE.md, the canonical end-user update
procedure. Motivated by a production deployment hitting timeouts
against a network device with a slow SNMP agent — defaults
preserve existing behavior, operators with similar issues can
raise the timeout via MNM_SNMP_TIMEOUT_SEC.

See UPGRADE.md for the deployment procedure. Tag-pinning is the
recommended flow for production installs.

---

Added

• MNM_SNMP_TIMEOUT_SEC and MNM_SNMP_RETRIES environment
  variables for tuning SNMP polling against slow devices.
  Defaults (10s, 1 retry) preserve existing behavior. Per-device
  overrides planned for v1.1.
• UPGRADE.md at the repo root documenting the canonical
  end-user update procedure. README links to it from a new
  "Updating" section.

Changed

• controller/app/polling.py now passes timeout_sec and
  retries to all SNMP collector calls
  (arp_snmp.collect_arp, mac_snmp.collect_mac,
  lldp_snmp.collect_lldp, snmp_collector.walk_table for
  the route SNMP fallbacks, and the collect_ifindex_to_name
  / collect_bridgeport_to_ifindex resolution helpers).
• Public collectors and snmp_collector.walk_table /
  get_scalar / collect_ifindex_to_name /
  collect_bridgeport_to_ifindex gained a retries: int = 1
  kwarg (default preserves existing behavior; CLI scripts and
  tests are unaffected).
• Discovery sweep (discovery.py:_snmp_get) retains its
  internal timeout=3, retries=1 values as a deliberate sweep
  optimization — documented inline. Folding sweep under the
  polling default would 3× dead-IP iteration cost (a /24 going
  from ~9 min to ~30 min with concurrency=16).
• Controller startup log now emits an effective SNMP polling
  configuration line so operators can see the active values in
  docker compose logs controller.