The shadow model: how and why small choices in spatially explicit species distribution models affect predictions