There should be only one transform per layer

[jni]

In his review of #40, @andy-sweet noted that every point in a points layer gets its own transform. This is wasteful since there is only one transform in a layer. Instead, all elements of one layer should be grouped in a <g> tag that contains the relevant transform. In the current implementation layer elements weren't grouped (they were all added to a single long list), but I think that was a flaw in the existing implementation, since svg editing software would allow operating on groups, and it might be desirable e.g. when making a figure to edit a whole layer together.