app_minus_test_thresh_weighted_sum: Weight feature importance sum on the basis of overfitting
In mdkessler/kTSCR: Implements k-Top Scoring Cluster Regression

When calculating feature importance by tallying the number of times a feature is in a sibling within and then across k-cv iterations, weight the tally by the overfitting seen in the iteration, as measured by app_cor - test_cor

app_minus_test_thresh_weighted_sum(
  k_cv_res,
  feature_importance_mat,
  k,
  penalty = 5
)

`k_cv_res`	result from run_cross_validation (called with condensed_output = TRUE)
`feature_importance_mat`	feature importance values from each kTSCR iteration derived from how many times a feature was seen in a sibling
`k`	k number of cross validation iterations, passed from run_cross_validation function
`penalty`	a penalty used in the weighting. A (potential) hyperparameter. Default is 5.