Between model: jaccard similarity

Jaccard similarity gives the proportion of matches in two lists.

- Produce Jaccard similarities for top 1%, 5%, etc. of predictions so we can see if the models are flagging the same people.

- Produce Jaccard similarities for top 1%, etc of feature importances so we can see if the models are using the same set of features

Start by locating the prediction-model evaluation entry point and the inputs for predictions and feature importances. Implement Jaccard similarities at the requested top 1%, 5%, and other stated cutoffs, then verify that results show overlap between models for both people flagged and features used.

Written by the indexing model from the issue text.