Implement some heuristics to sanity check the tumor and normal matching

[armish]

Things that can go wrong here include:

• Mismatched sequencing platform
• Sample mismatch/mix-up
• ...

And these can easily lead to some unexpected variant calling behavior (e.g. having tens of thousands of filter-passing somatic variants). Potentially, metrics extracted from the BAM file (and during the QC process) might help with implementing such heuristics — e.g. read length, # of reads, information content...