Random Forest regression, and out-of-bag error estimation

Implement:

def bootstrap_sample_with_oob(input, targets, seed=None) -> tuple[np.ndarray, np.ndarray, np.ndarray]:
    """Like 02-bagging's bootstrap_sample, plus the indices never drawn."""

def train_random_forest_regressor(input, targets, n_trees, max_depth, seed=None) -> list[tuple[dict, np.ndarray]]:
    """Returns [(tree, oob_indices), ...]."""

def predict_random_forest_regressor(forest, input) -> np.ndarray:
    """Averages every tree's prediction."""

def oob_error(forest, input, targets) -> float:
    """Mean squared error using only each sample's out-of-bag trees."""

• Reuse 05-regression-trees's build_regression_tree/predict_regression_tree.
• predict_random_forest_regressor averages (regression), it does not vote (01-random-forest-majority-vote was for classification).