How to generalize the first dimension (often assigned to batch size) on ONNX production inference?

### 📚 Documentation

Since we have to pass `input_sample` (`torch.randn((1, 64)`, for instance), when compiling the PL model to ONNX, we must pass only input with shape=(1,64), on inference time:

```python
# compiling the PL model to ONNX
model = PLModel()
filepath = "model.onnx"
input_sample = torch.randn((1, 64))
model.to_onnx(filepath, input_sample, export_params=True)

# inference time
import onnxruntime

ort_session = onnxruntime.InferenceSession(filepath)
input_name = ort_session.get_inputs()[0].name
ort_inputs = {input_name: np.random.randn(1, 64)}
ort_outs = ort_session.run(None, ort_inputs)
```

Then, how to generalize the first dimension (often assigned to batch size) on `onnx` production inference?

Start with the issue's model.to_onnx example and the ONNX Runtime InferenceSession example to identify the documentation entry point for production inference. Document how the first input dimension should be handled for variable batch sizes, including the expected export and inference behavior, and verify the example remains consistent.

Written by the indexing model from the issue text.