EnumeratedShapes converts with ClassifierConfig

## 🐞Describing the bug
Apparently, ClassifierConfig only supports batch size of 1, and the conversion will fail if you supply any other batch explicitly. However, is not the case with EnumeratedShapes - the converter passes just fine, but you will get an error when trying to run the model on a mac device.
I'd appreciate if the conversion script failed whenever the single-sample (B=1) native models encounter symbolic shapes.

## Stack Trace
```
input shape spec: EnumeratedShapes([(1, 32, 32, 3), (4, 32, 32, 3), (8, 32, 32, 3)], default=[4, 32, 32, 3])
Converting PyTorch Frontend ==> MIL Ops: 98%|█████████▊| 41/42 [00:00<00:00, 1658.45 ops/s]
Running MIL frontend_pytorch pipeline: 0%| | 0/5 [00:00

Start with the linked Colab reproduction and trace how EnumeratedShapes and ClassifierConfig are handled during conversion. Compare the behavior for batch size 1 and larger batches, then identify the conversion path for symbolic shapes. Done means unsupported native models fail during conversion rather than later on a Mac device, while supported inputs retain their current behavior.

Written by the indexing model from the issue text.