Restore deploy/manifest fixes lost when porting the optical controller onto develop

- opticalcontrollerservice: re-add the external NodePort (31060) Service
  used by REST clients and demo scripts outside the cluster.
- deviceservice: re-expose port 10065 (subcarrier_groups API). Without it,
  OpticalController's read_DSC_only() hung on a 10 s connect timeout on
  every compute-p2mp.
- OpticalController: Flask debug mode is now opt-in via
  OPTICAL_CONTROLLER_DEBUG (the Werkzeug debugger must not be reachable via
  the NodePort); read_DSC_only() uses a 2 s connect timeout.
- deploy/component.sh: extract the image tag with cut -f4 (the registry URL
  has a port, so -f3 never matched and upstream :latest images were
  deployed instead of the local build); create tmp/ folders; call
  ./deploy/show.sh instead of the non-existent ./show_deploy.sh.
- deploy/tfs.sh: fall back to Kafka port 9092 when kafka-service isn't
  ready, so the kfk-kpi-data secret never ships an empty port.
- scripts: restore run_tests_locally-telemetry-netconf.sh; make
  show_logs_opticalcontroller.sh follow the logs.

Verified: 4-cycle alarm failover/revert lifecycle test passes (28/28
checks) against device and opticalcontroller rebuilt from this tree.