Ideas for shortcut modelling

- [ ] Each shortcut has a detour ratio indicating likeliness. When we count shortcuts per road, we could try weighting by this.
- [ ] If we have good OD data for the predict-impact mode and have a sense of study area-wide traffic patterns, we could see what shortcuts currently occur in practice, and maybe weight the (entry, exit) pairs somehow?
- [ ] As you place filters, shortcut coloring constantly re-scales to show the worst problem. Optionally remember the original max_shortcuts before edits, and show relative to that?