Fix/cleanup

[swainsubrat]

Formalised and made the NIDS pipeline consistent to include these steps across all 6 NIDSs:
BaseSKLearnModel (LOF, OCSVM)
├── train_model(train_loader)
├── save(model_path)
├── load(model_path)
├── infer(test_loader)
└── evaluate(test_loader)

PyTorchModel (VAE, Autoencoder, ICL, KitNET)
├── train_model(train_loader)
├── save(model_path)
├── load(model_path)
├── infer(test_loader)
└── evaluate(y_test, y_pred, reconstruction_errors) # internal signature