Validate EfficientNetV2 MQTT inference in Docker

[KishanFW]

I updated the Engine inference flow to use the local EfficientNetV2 TFLite prediction path instead of the old TensorFlow Serving outputs format. The Engine now produces predicted species labels, confidence values, and top predictions from MQTT audio messages.

Files Changed

• light_echo_engine_efficientnetv2_tflite.py
  Updated MQTT audio inference to use EfficientNetV2 TFLite locally.

• system_manager.py
  Fixed simulator start mode handling.

• control_sim.py
  Updated simulator testing flow for Animal_Mode.

• comms_manager.py
  Added debug checks to confirm valid WAV audio was being sent.

• Engine requirements file
  Updated Engine dependencies needed for Docker/TFLite inference.

• Simulator requirements file
  Updated Simulator dependencies needed for MQTT/audio testing.

• README_EfficientNetV2_Engine_Integration.md
  Updated final Docker/MQTT validation status.

Note

The remaining Internal Server Error happens after prediction when sending results to the API/database, so it appears to be an API/backend issue.

[mraula]

everything looks good