stop_times: add shape_dist_traveled if it doesn't exist

[fpurcell]

Like #50, there is another (optional) distance value in GTFS. And to no surprise, RTD's gtfs is not populating those values. So as part of the post-process, we should also populate stop_time's shape_dist_traveled

Ideas:

• I guess we could do the 'as-the-crow-flies for stop-to-stop pairs too. That would be easier, and maybe a good first pass (and just good enough in general).
• a second idea would be to refer to the shape points to determine the stop_time distance (thinking this might be more accurate and more logically
• with this idea, we'd populate shapes first, then stop_time next, since stop_time will rely on those initial values (so this would be executed within shape's post process)
• the idea is to spatially match stop X to an entry in shape which matches both the trip's shape_id and also the nearest query. Thus we'd get our distance value. The problem here would be where a trip / shape visits a stop multiple times in the trip...would then be doing some gymnastics on which distance value to use--could be both slow and complicated.