Fix pedestrian routes using walkway_factor

[dgearhart]

Issue

We were seeing crazy routes when testing the walkway_factor values. After analyzing routes we did the following:

• Updated the walkway and sidewalk factor defaults
• Updated the AStarCostFactor method so we do not create crazy routes when favoring walkways
• Added more pedestrian test routes

BEFORE#1 walkway_factor = 0.1
https://github.com/valhalla/valhalla/blob/gdg_walkway_factor/test_requests/pedestrian_routes.txt#L4

AFTER#1 walkway_factor = 0.1
https://github.com/valhalla/valhalla/blob/gdg_walkway_factor/test_requests/pedestrian_routes.txt#L4

BEFORE#2 walkway_factor = 0.8
https://github.com/valhalla/valhalla/blob/gdg_walkway_factor/test_requests/pedestrian_routes.txt#L17

AFTER#2 walkway_factor = 0.8
https://github.com/valhalla/valhalla/blob/gdg_walkway_factor/test_requests/pedestrian_routes.txt#L17

Tasklist

• Test
• Review - you must request approval to merge any PR to master
• Generally use squash merge to rebase and clean comments before merging
• Update the changelog

[kevinkreiser]

noting that A*s goal estimation heuristic should never overestimate the actual cost to the goal, we landed on this implementation.

[kevinkreiser]

https://en.wikipedia.org/wiki/Consistent_heuristic <-- that is what i was referring to

[dnesbitt61]

You probably want to limit how small the factors can be. Making them very small values like 0.1 will force the A* factor to change and performance will drop. Also, it is usually more effective to penalize edges rather than favor them, so to dramatically favor walkways one might try penalizing regular roads (weighting factor > 1.0)

[dgearhart]

The favor examples were used to show that it was not working as expected prior to the change. The thinking is that it will be used more for avoiding every separate sidewalk to keep a simpler path and calling out street names for users