Qualia-plugin-snn

New features
- Bump maximum Python version to 3.12.
- `postprocessing.EnergyEstimationMetric`: print `ModelGraph`.
- `postprocessing.EnergyEstimationMetric`: trace `GlobalSumPooling` (not counted in metric).
- `learningmodel.pytorch.layers.quantized_SNN_layers.py`: membrane potential uses new tensor type `v` to enable quantizer configuration decoupled from activations.

Bug fixes
- `learningmodel.pytorch.SNN`: inherit from SpikingJelly's `StepModule` to avoid warning when using `set_step_mode()` on the network.
- `postprocessing.FuseBatchNorm`: copy `step_mode` and inherit returned `GraphModule` from SpikingJelly's `StepModule`.
- `pyproject.toml`: require `pytorch` dependency group of `qualia-core`.

Bug fixes

- QuantizedIFNode/QuantizedLIFNode: return correct round mode for membrane potential and hyperparameters for Qualia-CodeGen.

Bug fixes

- LearningModel SResNet/QuantizedSResNet Fix using 2D layers.
- PostProcessing EnergyEstimationMetric: Fix display of message when non-binary inputs/outputs are included in the total.

Other changes

- Some refactor for quantized layer inheritance.

Initial release of Qualia-Plugin-SNN.