Redesign of the logic for tumor only and germline samples

[jfnavarro]

The current approach is to perform variant calling (germline mode) for the samples even
if they are paired (tumor-normal). So, for instance if we have

SAMPLE1 0
SAMPLE2 1
SAMPLE3 1

Sarek will perform variant calling in somatic mode for SAMPLE2_vs_SAMPLE1 and in germline mode
for SAMPLE1, SAMPLE2 and SAMPLE3. I believe it is not necessary to perform variant calling
in germline mode for SAMPLE1 and SAMPLE2 (specially for SAMPLE1). By changing this behaviour
we would save a considerable amount of computational time and resources (specially for big datasets).

This, of course, may vary depending on the tools used. Any thoughts on this? Technically, it would not
require a lot of changes in the code to change the current behaviour.