Adds cucumber scenario for u turn restrictions at intersection

[daniel-j-h]

@srividyacb here's a cucumber scenario for what we talked about. See how the route does not make a u-turn at x but instead goes to the b node and makes a u-turn there.

[MoKob]

Currently this is blogged by #4465

[TheMarex]

We could use bearing values here to work around the limitation of snapping to x and trigger a real u-turn.

[SiarheiFedartsou]

@mjjbell as you worked on PR closing #4465 and this PR was blocked by it could you take a look? This test currently failing on master:

2022-10-06T14:33:17.0903046Z    Message:
2022-10-06T14:33:17.0903377Z      �[31mTables were not identical:
2022-10-06T14:33:17.0903835Z      | waypoints |     route             |
2022-10-06T14:33:17.0904218Z      | a,x,a     | (-) ax,ax,xb,xb,ax,ax |
2022-10-06T14:33:17.0904606Z      | a,x,a     | (+) ax,xb,xb,xb,ax,ax |�[39m

But it seems in this context ax,ax,xb,xb,ax,ax is the same as ax,xb,xb,xb,ax,ax, or not? Asking because tbh I don't have enough knowledge yet to now for sure if we can merge it or not(but probably would be great to merge it if it is ready) :)

[mjjbell]

Short answer: yes

Longer answer:
Adding the turns to see what's going on.

| a,x,a     | (-) ax,ax,xb,xb,ax,ax | (-) |
| a,x,a     | (+) ax,xb,xb,xb,ax,ax | (+) depart,new name straight,continue uturn,arrive,depart,arrive |

So the response is splitting the route into two as a->x->b(uturn)->x , x->a
whereas the test wants a->x , x->b(uturn)->x->a

They will have equivalent weight, so it's just whichever OSRM happens to select out of the two.

This wouldn't have worked in the past because the mid journey stop might have been on cx or somewhere else that wasn't optimal.