fix: use ONNX backend for torch2trt conversion in notebooks

[Mrudula-itsjuzme]

Resolves #178 by adding use_onnx=True to the torch2trt.torch2trt calls in the live_demo.ipynb and eval.ipynb notebooks. This prevents the __len__() should return >= 0 ValueError associated with the native torch2trt converter, ensuring the notebooks run out of the box on modern environments.