Different percentage for the same boarding, even with the same times

[lenaschimmel]

For some trips, the stop page shows a percentage, and the matching trip page shows another percentage. Of course, this might be the case when issue #6 occurs, but it also happens independently.

For example, this trip has the same timing information on both pages, but different percentages:

Stop page:

Trip page:

[lenaschimmel]

Some notes:

• The example above contains a walk, but the error is also reproducible on journeys without a walk
• On the trip page, we just show trip_data.start_prob
• On the stop page, we always add a walk (possible 0m) to the arrival and then compute the transfer probability to the departure, multiplied by the stop_data.start_prob
• The 0m-walk was intended to account for the time you need to change from one vehicle to the other, but:
  • we also compute it at the very start of the journey, where no such change event occurs
  • the current implementation of get_walk_time returns a linear distribution between -12s and 12s for walks shorter than 20m

[lenaschimmel]

As indicated by the previous comment, for transfers without an actual walk, the bug was caused by the 0m-pseudo-walk.

For actual walks, trip_page and stop_page use wildly different distances (e.g. 211m vs. 485m), and hence different walk times and probabilities.

This is a larger issue and needs redesign of our multi-stop handling. See #22 for this.