Multimodal_MLP_for_screen

Implementing half-up rounding for AUROC and PR AUC in the 'evaluate_cv_performance' function

Joining AUROC and PR AUC computation with summary table generation and computing mean and standard deviation of table metrics across k seeds/splits

Streaming inference results to disk instead of accumulating them in lists to prevent an OOM error

Also saving the best checkpoint based on validation AUROC

Modifying existing functions such that they process and plot multiple prediction files and also introducing a new function for k-fold cross-validation/k random seeds