add file to generate larger than memory datasets

[r-brink]

@ritchie46 we discussed it last week. Not sure if what you want to do with this code, so I'll just offer it as PR.

We could also merge both prepare_files scripts. Then users can generate the datasets no matter the scale factor.

[ritchie46]

Ah nice one. For when we go biiiig!