More realistic precision for cosine distance results

Chris Veness' discussion of using the spherical law of cosines (http://www.movable-type.co.uk/scripts/latlong.html) suggests it gives well-conditioned results "around 1 metre" in JavaScript (this assumes a sphere with 6,371km radius).

In Chrome 33, Math.pow(Math.cos(Math.PI / 4), 2) yields 0.4999999999999999.  When we take the arccosine of twice this, we get something significantly different than zero.  Multiplying by 6371 means we can't assert that this is within 1e-9 of zero.

test/spec/ol/sphere/sphere.test.js

@@ -379,7 +379,7 @@ describe('ol.Sphere', function() {
       for (i = 0; i < expected.length; ++i) {
         e = expected[i];
         expect(sphere.cosineDistance(e.c1, e.c2)).to.roughlyEqual(
-            e.cosineDistance, 1e-9);
+            e.cosineDistance, 1e-3);
       }
     });