Instead of having every drawable object know how to transform itself based on
the camera position, we pass in a transformation matrix to spriteBatch.Draw().
Unfortunately MonoGame only lets us specify a translation that works over an
entire SpriteBatch.Begin() call, so we need to begin & end separately for
objects that *aren't* supposed to translate at the same rate as the camera.
Fixes#39.
GitOrigin-RevId: afab72c39236b1b46fe1597412209981ddae9c7c