Update TensorRT repository branch to match TensorRT installed

[kgonia]

I tried running the Stable Diffusion example; however, there is a problem with running it due to a TensorRT version mismatch.

The container has installed version 10+. Here is the result of pip list:

tensorrt 10.0.1 tensorrt-cu12 10.0.1 tensorrt-cu12-bindings 10.0.1 tensorrt-cu12-libs 10.0.1 

But Docker refers to a repository with version 9.

There was a change in the API:

https://docs.nvidia.com/deeplearning/tensorrt/migration-guide/index.html

get_binding_dtype is now get_tensor_dtype.

With version 9, I got this error:

tritonserver.InvalidArgumentError: load failed for model 'stable_diffusion_1_5': version 1 is at UNAVAILABLE state: Internal: AttributeError: 'tensorrt_bindings.tensorrt.ICudaEngine' object has no attribute 'get_binding_dtype' At: /opt/tritonserver/backends/diffusion/Diffusion/utilities.py(225): allocate_buffers /opt/tritonserver/backends/diffusion/Diffusion/stable_diffusion_pipeline.py(253): loadResources /opt/tritonserver/backends/diffusion/model.py(140): initialize ;