Add Default impl for Dataset to avoid writing out all empty fields

I not sure whether this is something that we should do. For one thing, Dataset does not have a meaningful default value even if it is technically well-defined. But more importantly, while this does save a lot of typing for harvesters which only fill in a few fields (which might stay the norm for more obscure sources) especially as new fields are added, it might actually be good to have to write out these fields to more easily see where improvements in coverage are possible.

Hence the separate MR for this relatively small change so we can decided whether this is helpful or not...