Remove VehicleToStopHeuristics

[leonardehrenfried]

Summary

I received several reports that this sandbox feature, which I wrote quite a while ago, was returning very strange results.

Since we have a much better way to control the access/egress with either a maxStopCount or the penalty, rather than investigating why it fails, I decided to remove it

Fixes #5337

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (14aca8a) 67.54% compared to head (d06a4db) 67.63%.
Report is 1 commits behind head on dev-2.x.

Additional details and impacted files

@@              Coverage Diff              @@
##             dev-2.x    #5381      +/-   ##
=============================================
+ Coverage      67.54%   67.63%   +0.09%     
- Complexity     16293    16297       +4     
=============================================
  Files           1888     1886       -2     
  Lines          71609    71570      -39     
  Branches        7405     7390      -15     
=============================================
+ Hits           48365    48407      +42     
+ Misses         20741    20661      -80     
+ Partials        2503     2502       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[t2gran]

Should we add default time-penalty config before removing this ? That would help people using it, since the system will behave almost the same.

[leonardehrenfried]

I think that's a good idea.

[vesameskanen]

This is otherwise OK, but the new fairly high default penalties create a side effect, which is sort of a known problem: penalty is applied only when the defined street modes are changed to a transit mode. This causes a strange bias, because transit gets penalty whereas another street mode does not. So, walking several kilometers after a car leg is cheaper than taking much faster transit connection. See attached picture. This needs a solution - any ideas (other than configuring penalties back to zero)?

[leonardehrenfried]

This PR has come back to life after a long sleep. And I fixed the unfair advantage of direct results in #5483 so this is ready for review again.

@vesameskanen @vpaturet Can you take a look at it?

[vpaturet]

@t2gran this feature is still in use at Entur. What would be the proper way to update our configuration?

[leonardehrenfried]

@vpaturet This PR now sets a default accessEgress penalty for the car modes. They will be good choices for the majority of deployments, including Entur, so you don't need to do anything.

With your expertise, however, you might want to finetune the values.

[t2gran]

@vesameskanen Are the default settings too aggressive?

[vesameskanen]

The default penalty values look high, but current park&ride itineraries seem to be sensible. Our service makes a separate search for plain park & ride, with a high car penalty value of 10.

[vesameskanen]

I tested much lower penalty (5m + 1.2t, 1.2). That seems to add variation to itineraries but also adds more car driving. Finding the optimal level is not easy. We will probably use the default values.