iibw

Phi 3 vision support exists in [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/multimodal#phi-3-vision), but not in tensorrtllm_backend. Could support be added to tensorrtllm_backend now that we have some multimodal support for Blip2 and Llava? Thanks for...