Add LayerNorm support

[gcunhase]

Issue

Failed to convert ONNX model with LayerNormalization layer (opset>=17):

File "onnx2pytorch/onnx2pytorch/convert/operations.py", line 287, in convert_operations
    raise NotImplementedError(
NotImplementedError: Conversion not implemented for op_type=LayerNormalization.

How to repro

Run:

# 1. Export simple ONNX model with LayerNorm
import torch
from torch import nn
block = nn.LayerNorm([128])
input_tensor = torch.randn(1, 32, 128, 128)
onnx_path = "model_ln.onnx"
torch.onnx.export(
    block,
    input_tensor,
    onnx_path,
    input_names=['input'],
    output_names=['output'],
    opset_version=17
)

# 2. Convert ONNX model back to PyTorch
import onnx
import onnx2pytorch
onnx_model = onnx.load(onnx_path)
pytorch_model = onnx2pytorch.ConvertModel(onnx_model)  # -> Error occurs here

# 3. Recover ONNX model and check they're the same
torch.onnx.export(
    pytorch_model,
    input_tensor,
    onnx_path.replace(".onnx", "_recovered.onnx"),
    input_names=['input'],
    output_names=['output'],
    opset_version=17
)

[gcunhase]

Debugging error with dynamic calculation in LN op, moving this to draft for now.

[gcunhase]

Fixed, ready for review.