pub fn evaluate_counts( ground_truth: &[usize], predictions: &[usize], ) -> Result<CountingMetrics, EvalError>