fix(antenna worker): use classifier.detector_cls instead of hardcoded APIMothDetector

The original detector_cls refactor in #141 updated the API path (api.py)
but missed the antenna worker path. When a pipeline declares a non-default
detector (e.g. mothbot_insect_orders_2025 with YOLO11m), the worker was
still running FasterRCNN detection, then posting results with the FasterRCNN
algorithm key, and Antenna rejected them because the /info endpoint declared
only the YOLO and order-classifier algorithms. NATS re-delivered the message
in a loop.

Discovered when testing job 1492 on the local Antenna at beast:
  Error processing pipeline result for job 1492: Detection algorithm
  fasterrcnn_for_ami_moth_traps_2023 is not a known algorithm. Known:
  ['insect_order_classifier_mothbot_yolo_detector',
  'mothbot_yolo11m_creature_detector'].

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: mihow

trapdata/antenna/worker.py

@@ -464,7 +464,7 @@ def _process_job(
             # Defer instantiation of poster, detector and classifiers until we have data
             if not classifier:
                 classifier = classifier_class(source_images=[], detections=[])
-                detector = APIMothDetector([])
+                detector = classifier_class.detector_cls([])
                 result_poster = ResultPoster(max_pending=MAX_PENDING_POSTS)
 
                 if use_binary_filter: