proper tumor sample id identification with different sample classes

[covingto]

At the moment the system attempts to isolate the Normal and Primary samples from the header of the VCF files. We find that some files use not a Primary sample but rather a Metastasis sample. This relates to the block beginning at line 137 of merge.py.

We may propose a few alternatives including adding logic that ID=METASTASIS be added as a valid primaryline id. But there may be other cases that require a finer grain of sample identification.