Unit-tests are not flagging models with incompatible output structure

The model `aguilara42/audacity-Wav2Vec2-Base` is causing the following assertion error:
```
~/Desktop/audacity/modules/mod-deep-learning/DeepLearningAnalyzer.cpp(193): assert "output.dim() == 1" failed in TensorToLabelTrack().
```

This should be flagged during the serialization process. See also audacitorch/audacity#90.