Stretch: bagging concept

Implement:

def bootstrap_sample(input: np.ndarray, labels: np.ndarray, seed: int | None = None) -> tuple[np.ndarray, np.ndarray]:
    """Resample n_samples rows with replacement."""

def train_random_forest(input, labels, n_trees, max_depth, seed=None) -> list[dict]:
    """Train n_trees independent trees, each on its own bootstrap sample."""

• Reuse 03-best-split-minimal-tree's build_tree, don't reimplement tree construction.
• The returned list is directly usable by 01-random-forest-majority-vote's random_forest_predict.
• Build exactly one random generator, before the loop over trees, not one per tree.