Make data "messier"

> make the random cdisc data (ideally the cached one) much much messier as that will help the core team:

a) understand what sorts of messiness we're likely to see - trying to catch everything is unlikely to be efficient but we can catch common things (e.g. I had no idea ARM and ACTARM could have different contents)

NEST/teal.goshawk/pull/217#issuecomment-69891_

Provenance:
```
Creator: anajens
```