Clockwise fix

[flippmoke]

Fixed the way the decoder dealt with discarding interior and exterior rings when using bounding boxes. Also moved processing of vector tile area into vector tile coordinates so that clockwise and counterclockwise are more properly calculated.

/cc @springmeyer @jakepruitt

[artemp]

@flippmoke - why are you not using mapnik::util::is_clockwise? We must stamp down on duplicate code as it's only encourages 🐛 's .

[flippmoke]

@artemp it no longer does it in the same way as it is done in a streaming fashion as the geometry is decoded. Therefore, I can no longer use the same method.

[flippmoke]

Current Master:

./build/Release/vtile-decode ./test/data/0.0.0.vector.mvt 0 0 0 100000
z:0 x:0 y:0 iterations:100000
message: gzip compressed
8313.71ms (cpu 8282.74ms)   | decode as datasource_pbf: ./test/data/0.0.0.vector.mvt
processed 800000 features

Pull Request:

./build/Release/vtile-decode ./test/data/0.0.0.vector.mvt 0 0 0 100000
z:0 x:0 y:0 iterations:100000
message: gzip compressed
7916.56ms (cpu 7883.13ms)   | decode as datasource_pbf: ./test/data/0.0.0.vector.mvt
processed 800000 features

[springmeyer]

@flippmoke:

1. What is the reasoning for additional else in general? E.g. why set to empty vs what was it before?
2. I think the std::move is useless here since mapnik::geometry::geometry_empty() is already a temporary and it can be just geom = mapnik::geometry::geometry_empty(); /cc @artemp for confirmation.

[flippmoke]

I believe it actually is useless, in my mind I always wanted geom set in some manner in the code but then I remembered the default initialization for mapnik::geometry:geometry<> is mapnik::geometry::geometry_empty

[springmeyer]

@flippmoke - this looks great, I think it is ready to 🚢