Model caching

[pjcozzi]

Would like to get this in for 1.5.

Open the Chrome task manager, and note the memory usage and load time for the following Sandcastle example using both master and this branch.

var viewer = new Cesium.Viewer('cesiumContainer');
var scene = viewer.scene;

function createModel(url, position, zoom) {
    var modelMatrix = Cesium.Transforms.eastNorthUpToFixedFrame(position);

    var model = scene.primitives.add(Cesium.Model.fromGltf({
        url : url,
        modelMatrix : modelMatrix,
        scale : 0.25 + (1.75 * Cesium.Math.nextRandomNumber()),
        minimumPixelSize : 128
    }));

    model.readyToRender.addEventListener(function(model) {
        // Play and loop all animations at half-speed
        model.activeAnimations.addAll({
            speedup : 0.2 + (2.0 * Cesium.Math.nextRandomNumber()),
            loop : Cesium.ModelAnimationLoop.REPEAT
        });

        // Zoom to model
        if (zoom) {
            var center = Cesium.Matrix4.multiplyByPoint(model.modelMatrix, model.boundingSphere.center, new Cesium.Cartesian3());
            var transform = Cesium.Transforms.eastNorthUpToFixedFrame(center);
            var camera = scene.camera;
            camera.transform = transform;
            var controller = scene.screenSpaceCameraController;
            var r = 2.0 * Math.max(model.boundingSphere.radius, camera.frustum.near);
            controller.minimumZoomDistance = r * 0.5;
            camera.lookAt(new Cesium.Cartesian3(r, r, r), Cesium.Cartesian3.ZERO, Cesium.Cartesian3.UNIT_Z);
        }
    });
}

//var url = '../../SampleData/models/CesiumAir/Cesium_Air.gltf';
//var url = '../../SampleData/models/CesiumGround/Cesium_Ground.gltf';
 var url = '../../SampleData/models/CesiumMan/Cesium_Man.gltf';

var zoom = true;
for (var x = 0; x < 20; ++x) {
    for (var y = 0; y < 20; ++y) {
        createModel(url, Cesium.Cartesian3.fromDegrees(-123.0 - (x * 0.0002), 44.0 - (y * 0.0002), 0.0), zoom);
        zoom = false;
    }
}

Master:

GPU Process: 2.7 GB
Tab: 680 MB

This branch:

GPU Process: 323 MB
Tab: 400 MB

This adds caching to glTF models. This is a replacement for #2177. Although this doesn't include any commits from that pull request, I have still credited @Relfos in CONTRIBUTORS.md for his significant work here.

Also added a few ideas for improvements to the model roadmap: #927

[pjcozzi]

@bagnell can you review?

[pjcozzi]

@bagnell we should not have let these TODOs get into master (they are on the roadmap). I'll remove them shortly in my next pull request.

[bagnell]

OK. I wasn't sure because we have other TODO, PERFORMANCE_IDEA, PERFORMANCE_TODO comments. Some of which I implemented optimizing some of the geometries.

[pjcozzi]

Sometimes I leave a PERFORMANCE_IDEA (and they even come in handy once in a while), but call me on it if I leave a TODO in.