Using SSE SIMD in Simplification

[flippmoke]

A speed improvement in the simplification code by using SSE code.

Master Bench

./build/Release/vtile-encode ./test/geometry-test-data/benchmark/ginormous-polygon.json 0 0 0 -i 100 -p 4326
z:0 x:0 y:0 iterations:100
5421.31ms (cpu 5407.38ms)   | encode tile: ./test/geometry-test-data/benchmark/ginormous-polygon.json

Pull Request Bench

./build/Release/vtile-encode ./test/geometry-test-data/benchmark/ginormous-polygon.json 0 0 0 -i 100 -p 4326
z:0 x:0 y:0 iterations:100
3009.47ms (cpu 2998.82ms)   | encode tile: ./test/geometry-test-data/benchmark/ginormous-polygon.json

/cc @springmeyer @artemp @jakepruitt

[pnorman]

Will this continue to work on non-x86 platforms?

[flippmoke]

@pnorman if we decided to accept this pull request we could add a compile flag to use the original functionality. It should work just fine on x64 as well

[pnorman]

cc @sebastic since I think he's packaging mapnik-vector-tile on other CPUs

[sebastic]

The mapnik-vector-tile source package builds an architecture independent binary package, so the build is only performed on the architecture of the build system which is amd64 in my case. The node-mapnik package may be affected by this if it's arch:any, it's the reason we have mapnik-vector-tile packaged.

[flippmoke]

After latest commit:

./build/Release/vtile-encode ./test/geometry-test-data/benchmark/ginormous-polygon.json 0 0 0 -i 100 -p 4326
z:0 x:0 y:0 iterations:100
2959.72ms (cpu 2936.32ms)   | encode tile: ./test/geometry-test-data/benchmark/ginormous-polygon.json

[pnorman]

This PR seems oddly large, like a merge or rebase screwed up

[flippmoke]

@pnorman a screw up in a rebase

[flippmoke]

Closed in favor of #268