OSM data parsing should retain OSM vertex id

[esaule]

The way we currently parse OSM data, we destroy the information of the OSM vertex ID by creating our own internal ID map.
This saves a little bit of data, but fundamentally prevent stitching maps together. The client should be written to retain OSM vertex IDs and use that for vertex identification.

We should check what the other clients are doing.

[esaule]

Fixed upstream in the C++ client. Should be tested and checked on the other two clients.