E3 detectors do not consider time step fragments when computing meanDurationWithin

[m-kro]

The bug can be observed in the test sumo/output/e3/one_vehicle/one1s_1vehicle_10m_s. A single vehicle passes the detector during the simulation. It needs 20.3s to pass the detector as printed correctly in the output attribute meanOverlapTravelTime, but the attribute meanDurationWithin reaches a maximum of 21s.

[behrisch]

also a topic for the traffic tower session on detectors and timing