feat: add discount price dataset preparation and prediction models

[ShivamSaini09]

Summary

Adds dataset preparation and model training notebooks for next-week discount-price prediction.
Changes

• Cleans OCR-derived prices and standardises product names.
• Creates weekly product histories with price, promotion, and seasonal features.
• Creates targets for next-week special status and discount percentage.
• Compares classification and regression models, including XGBoost and a neural network.
• Combines the selected models to predict:
  • Special probability
  • Discount percentage
  • Expected special price
• Rounds predicted discounts to the nearest 5%.

Validation

• 6,784 products and 45,431 product-week rows retained.
• 44,640 classification examples.
• 2,376 trustworthy discount-regression examples.
• HistGradientBoosting achieved a test MAE of approximately 9.14 percentage points.

Notes

CSV datasets and generated model files remain local and are excluded by Git ignore rules. The source catalogue CSV is required to run the notebooks.

[lamtrvu]

This PR is ready to merge.