Fix the bug: tensorrt-cpp input blob from BGR to RGB

[AllenIsaacJose]

Opencv imread operation reads the image in BGR format. So when creating the input blob, we should make sure it is in RGB format before inferencing on the network. I had a lower accuracy in the existing code and this conversion of input blob to RGB format improved the accuracy.

[CLAassistant]

All committers have signed the CLA.