[RFC] Add torchvision/faster-rcnn inference example on cpp backend

Nobody has claimed this yet.

**Is your feature request related to a problem? Please describe.**

[TorchVision](https://github.com/pytorch/vision) have supported `faster-rcnn` and `retinanet`, refer to [here](https://github.com/pytorch/vision/blob/7b9d30eb7c4d92490d9ac038a140398e0a690db6/test/test_onnx.py#L366), this two models can be export to `onnx` and infer with `onnxruntime` python backend. I wanne know whether there is a plan to add an inference example of this two models on cpp backend?

**System information**
- ONNX Runtime version (you are using): build from master branch

**Describe the solution you'd like**

Add an inference example of torchvision's `retinanet` or `faster-rcnn` on C++ backend

Start with the linked TorchVision test/test_onnx.py reference for faster-rcnn and retinanet, then review how ONNX Runtime represents C++ inference examples. Decide which model the example should cover and what the C++ backend entry point requires; done means the selected model has a working C++ inference example.

Written by the indexing model from the issue text.