GRU support

[ZM-J]

There's a GRU onnx model which can be directly used to infer using onnxruntime. However, when I'm trying to use the following code to convert this model to its pytorch counterpart:

import onnx
from onnx2pytorch import ConvertModel

onnx_model = onnx.load(onnx_model_path)
pytorch_model = ConvertModel(onnx_model)

import torch
X = torch.randn((1, 60, 78))
with torch.inference_mode():
    y = pytorch_model(X)

print(y.shape)

I got the following error:

Traceback (most recent call last):
  File "/convert.py", line 41, in <module>
    y = pytorch_model(X)
  File "/miniconda3/envs/convert_model/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1736, in _wrapped_call_impl
    return self._call_impl(*args, **kwargs)
  File "/miniconda3/envs/convert_model/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1747, in _call_impl
    return forward_call(*args, **kwargs)
  File "/miniconda3/envs/convert_model/lib/python3.10/site-packages/onnx2pytorch/convert/model.py", line 224, in forward
    activations[out_op_id] = op(*in_activations)
TypeError: gru() received an invalid combination of arguments - got (Tensor, Tensor, Tensor, Tensor, Tensor, Tensor), but expected one of:
 * (Tensor data, Tensor batch_sizes, Tensor hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional)
 * (Tensor input, Tensor hx, tuple of Tensors params, bool has_biases, int num_layers, float dropout, bool train, bool bidirectional, bool batch_first)

I don't know why it has happened internally though :(