Routing result simplification

[farfromrefug]

This PR changes the points for the RoutingResult.
Before we were creating the points based on the manoeuvres. However this was adding unwanted duplicate points(multiple manoeuvres at the same point?).
On an example route (computed with android dev here):

• points returned from valhalla : 246
• points in returned RoutingResult: 350

Now with that PR it returns the shape returned from valhalla. And the instruction indexes are still correct.

[mtehver]

I could be wrong but I think the 'pointIndex' calculation is correct only if there is a single leg in the result. A simple check like below should remove redundant points and work in case of multiple legs also:

                    for (std::size_t j = static_cast<std::size_t>(maneuver.get("begin_shape_index").get<std::int64_t>()); j <= static_cast<std::size_t>(maneuver.get("end_shape_index").get<std::int64_t>()); j++) {
                        const valhalla::midgard::PointLL& point = shape.at(j);
                        MapPos epsg3857Point = epsg3857.fromLatLong(point.second, point.first);
                        if (epsg3857Points.empty() || epsg3857Points.back() != epsg3857Point) {
                            epsg3857Points.push_back(epsg3857Point);
                            points.push_back(proj->fromLatLong(point.second, point.first));
                        }
                    }

[farfromrefug]

@mtehver yes you are right i suppoed there aws only one leg! Do you know in which case we have multiple leg so i could try?

[farfromrefug]

@mtehver i finally took a look at that and i tired what you suggest and it did not work. Not sure exactly why but i still had all the points.
Now i took a second though about it and the fix was pretty easy. To ensure the point index was good with multiple shapes i simply had to do pointIndex=points.size()+maneuver.get("begin_shape_index").get<std::int64_t>()