Mapping columns to categories

In the current nyc_taxi.ipynb, pickups vs. dropoffs are aggregated separately, because each of them provides a set of coordinates yet aggregation can accept only a single coordinate per row. We should show how to use Pandas to restructure the data so that pickups and dropoffs appear as separate rows, with the type of event coded as a category ('p' vs. 'd', for instance), akin to how race is shown for the census example. This will let us work directly with the pickup and dropoff information per pixel, e.g. to allow intermediate colors showing mixing of the categories.