[Routing] Improving quality of OSM data for routing.

[Nic Roets]

I moved the data access functions of gosmore into a seperate cpp and h
file / library. The new API is very crude but allows bbox scanning,
text searching and routing queries.

So as an application, I wrote a program (osmunda) that scans a GPX
tracklog for maneuvers that are 'impossible' according to the given
OSM data. Specifically, it takes any two consecutive tracklog points
and calculates the route according to OSM data between the two. If
this route includes one or more complete segment, it flags the two
tracklog points by writing them to another GPX file.

It processed my test file of 14000 points in 8 seconds.

It tends to falsely flag dual carriage ways as the GPS receiver
occasionally generates points that are closer to the oncoming way. It
may also flag bridges, as it's unsure on which level the person was.
These can be reduced by adding code that considers the motion vector.

It also falsely flags movements through open areas, like parking areas
and city squares, as they will simply not match the routable ways.

But it did highlight one footway and one service road this hasn't yet
been added to the map.

Regards,
Nic