correct intersections for named area (explicitly check them against t…

[patrickbr]

…he R-Tree), this addresses point 1 from issue #14

[lehmann-4178656ch]

Looks good, just a quick question: does this change anything in the expected output - or better: can this change be represented via a test to ensure I don't break it in the future?

[patrickbr]

Yes, before this commit, an "intersects" was only written for named areas where one contained the other, so the output changed quite a bit (many more intersects relations). It should be easy to add a test for this, I will look into this :)

BTW, we still don't have way/way intersects, but this would require having random access to the ways stored on disk (to retrieve the exact geometry for intersect candidates). I am not sure whether the current boost::serialization approach can handle random accesses (maybe it can).

[patrickbr]

@lehmann-4178656ch, the named area intersections are now tested. I also made the intersect relations symmetric, that is, if a intersects_area b is added, b intersects_area a is also added (the same for intersects_nonarea).