feat: 🎸 version 0.4.0

[zezhishao]

• 1. Modify data processing: standardize data formats, add metadata, refactor dataset classes, and remove data normalization operations.
• 2. Add a Scaler class: manage data normalization and denormalization in an extensible way.
• 3. Delete the loss folder and merge its contents into metrics.
• 4. Modify the Runner: streamline training and inference interfaces, add post-training inference functionality, and add the ability to save inference results and evaluation metrics.
• 5. Add an examples folder: provide all configuration options with explanations, typical configuration files, and an example model structure (MLP).
• 6. Modify training and inference scripts.
• 7. Add a detailed tutorial.
• 8. Refine code details.
• 9. Update requirements.txt.
• 10. Update README.md.