Add TensorFlow Lite and Google Edge TPU support

[petertorelli]

Two notes:

• The Edge TPU was tested on the Coral Dev Board, ready to upload scores;
• TFLite in larger systems has significant performance or compatibility issues based on the TensorFlow or tflite_runtime versions. Releasing this for educational purposes only
• New TFLite fp32 and int8 models released, all are derived from the "golden" models in MLMarks "tensorflow" model area. The int8 models are quantized with PTIQ strategy, refer to the README file in the model area for the code used for the conversions.