Changes and new features in JeVois 1.17

JeVois 1.17 brings the following new features:

• Updated to OpenCV 4.5.4, OpenVino 2021.4 LTS
• Updated to TensorFlow 2.5.0
• Threadpool improvements, with better big/little affinity handling on JeVois-Pro.
• JeVois-A33: New JeVois DNN framework to run a variety of deep learning models for classification, detection, and semantic segmentation using OpenCV (supports Caffe, ONNX, TensorFlow, etc). Add your own models to JEVOIS:/share/dnn/models.yml (and also copy the corresponding model and weight files to microSD) to run them directly.
• DNN: friendlier error messages when expected network inputs/outputs differ from what is received.
• JeVois-Pro: minor linux kernel enhancements, updated NPU kernel module.
• Upgraded to latest numpy, which is now installed by default on platform, with corresponding API changes.
• JeVois-Pro: Better build and packaging system.
• JeVois-Pro: More deep learning models in models.yml model zoo.
• JeVois-Pro: Better fan control.
• Added more docs.
• Miscellaneous bug fixes and performance improvements.