Task: Update `dev_rank` to be based on the standard KDE.py calculation

## Describe the task
The `dev_rank` associated columns are currently based on aggregations to Census block groups. It would be simpler and more statistically valid to have them be based on the z-scores of building permits in the area, much like the other KDE approaches that we've used for crime and L&I complaints. Update the `dev_probability` function in `new_etl` to use the KDE approach. Be aware that there are some issues with the Carto query for this dataset, which is why the current function is written the way it is. You'll have to energy the function correctly add some kind of handling in `featurelayer.py` or `kde.py` to account for this situation.