PyTorch's BatchNorm2d causes script conversion error

PyTorch's *BatchNorm2d* in
https://github.com/pytorch/pytorch/blob/master/torch/nn/modules/batchnorm.py
causes issues on *PyTorch 1.9.1* - *1.11* with *Python 3.8* and *CoreMLTools 5.2.0*.

```
import coremltools as ct
import torch
from torch import nn, Tensor
import sys

class MyBatchNormModule1(nn.Module):
def __init__(self):
super().__init__()
self.bn = nn.BatchNorm2d(100, affine=False)
def forward(self, x: Tensor)->Tensor:
res: Tensor = self.bn(x)
return res

if __name__ == '__main__':
t1 = MyBatchNormModule1()
t1.eval()
x = torch.rand(20, 100, 35, 45)
t = torch.jit.script(t1)
result = t1(x)
model = ct.convert(t, inputs=[ct.TensorType(shape=x.shape)], convert_to='mlprogram', debug=True)
sys.exit(0)
```
The error is an `AssertionError: assert len(cond) == len(node.outputs)` in *converters/mil/frontend/torch/ops.py*
https://github.com/apple/coremltools/blob/main/coremltools/converters/mil/frontend/torch/ops.py

```
@register_torch_op(torch_alias=["if"])
def _if(context, node):
```

Run the supplied Python reproducer with the listed PyTorch and CoreMLTools versions, then inspect coremltools/converters/mil/frontend/torch/ops.py at _if and the len(cond) assertion. Done means the BatchNorm2d affine=False scripted model converts to an mlprogram without the reported AssertionError.

Written by the indexing model from the issue text.