Reconcile vehkm table with legs table

[Theodore-Chatziioannou]

Vehicle-km (for each car user) figures can be calculated from the legs table, or taken directly from the intermediate vehkm output. Testing with one of our sims, I get slightly different numbers in the two tables. This includes agents with zero veh-km travelled in the legs log, but non-zero distance in the vehkm log.
I think this might be due to different treatment of entry/exit links. I believe the MATSim assumption (reflected in the legs table) is that a vehicle enters "just before" the end of the link and exits "just after" its start, while the Gelato vehkm calculation is assigning the full length every time. Similarly, this affects any zero-distance trips: a single-link event with zero travel time is assigned the full distance of the link.