To fix bug 1007438 we need a way to get a measure of the memory used by the DrawTarget that is stored in VectorImages' SurfaceCache entries. I'm not sure why we're storing a DrawTarget rather than a SourceSurface, so probably we should have a way to get a measure of the memory used by by a SourceSurface too. If fact we'll probably want different methods to get the usage of the different types of memory that can be involved (something along the lines of what mozilla::image::Image provides with OutOfProcessSizeOfDecoded et. al.).