Add residualized outcomes and corrected heterogeneity tests

[jiongyi-cao]

• Updates the fixed-score and cross-validated heterogeneity tests using the corrected covariance estimators.
• Adds automatic construction of the residualized pseudo-outcome before testing.
• Uses SuperLearner with five-fold cross-validation by default, while allowing users to specify a different SuperLearner library or provide their own outcome model.