Allow onnx kwargs

[lazToum]

Example usage:

model_trt = torch2trt(
    model,
    [inputs],
    max_batch_size=max_batch_size,
    use_onnx=True,
    onnx_kwargs={'opset_version': 13, 'do_constant_folding': True}
)