Ensure that mutations are from only one tumor per patient

[petermchale]

Our model assumes that the probability of generating a mutation at a given site in a given tumor is independent of whether a mutation has been seen at the same site in another tumor. That assumption is violated when tumors share a common ancestor, as is the case for a primary tumor and a metastasis. Therefore one would probably want to ensure that no more than one tumor per patient is used when collecting SNVs.