TFLite and Pytorch CPP launcher, fps measure change and minor updates
Created by: maslovaz
- change FPS metrics calcuation: fps = overall num of images inferred / overall time of inference for all python launchers
- add batch fps metrics: batch_fps = batch size / median_latency
- add json report to all python launchers
- add pytorch c++ launcher
- add tensorflow Lite c++ launcher
- move latency calculation to dedicated decorator