More sophisticated RTree implementation

[ahocevar]

This new implementation is based on
http://github.com/imbcmdth/RTree/, with only a few modifications
to add the optional type and provide the API of the previous
implementation.

There is still room for optimization, but this is such an
improvement over the previous RTree already that it's worth
bringing it in now.

[ahocevar]

More work needs to be done to make this work in advanced compilation mode.

[ahocevar]

This is ready for review. I'm sure the Closure integration can be improved, but for now it works.

[tschaub]

Thanks for taking this on @ahocevar. Looks like a few more changes are needed to satisfy JSDoc3. I hope to get a chance to look this over in the next day or two.

[ahocevar]

@tschaub: JSDoc3 is also happy now. Don't worry if you don't see anything in the resulting docs - that's ok because there is nothing exported to the public API.

[tschaub]

Minor, but it would match the rest of our style to have braces here (and below).

[tschaub]

This is a fantastic improvement @ahocevar. Thanks for taking the time to adapt it.

While it's a silly example, its nice to see the canvas renderer handling 20K features with aplomb: https://gist.github.com/tschaub/5617621

On my machine, these synthetic features take about 1s to create, about 1s to index (with 266ms in goog.array.concat), and under 400ms to render (~130ms in ol.renderer.canvas.VectorRenderer.renderPointFeatures_, ~90ms in ol.style.Shape.createLiteral, ~85ms in ol.layer.Vector.groupFeaturesBySymbolizerLiteral). And there is plenty of room for optimization (for example, for simple cases of a single symbolizer literal, we can check symbolizer.isLiteral() and avoid all the time spent iterating through features, creating literals, and then grouping). It's clear that time spent loading real data will quickly eclipse this rendering time.

I can imagine we'll have discussion about where this gets used. It may only be of use in the Canvas 2d rendering case (in which case it should not be required by the vector layer). But I don't think that should keep this from going in.

Please merge.