Implement two common outlier rules, each returning a boolean mask:
iqr_outliers(x, k=1.5): outside [Q1 - k·IQR, Q3 + k·IQR], with quartiles from np.percentile (default linear interpolation)zscore_outliers(x, threshold=3.0): |x - mean| / std > threshold (population std). A constant array has no outliers.