Add reproducible k-fold splitting and CSV export to Bagging

[rizoudal]

This pull request introduces two related features to improve the reproducibility of k-fold cross-validation in the Bagging class:

1. Seed parameter

   • Adds a seed class attribute to allow deterministic behavior in sampling_kfold.

2. CSV export of splits

   • Saves the generated train/val splits for each fold to a CSV file (kfold_split.csv).
   • Includes fold number, subset type (train/val), sample name, and the class label.

Let me know if anything should be adjusted.

[muellerdo]

Nice!