improve variant calling pipeline

[k8hertweck]

• add step to remove PCR errors (after read mapping, before SNP calling), using Picard MarkDuplicates
• filter out sites with extremely high depths of coverage (put vcf file into R, make distribution of depths, remove any sites that are 1.5 - 2x the median depth of coverage to get rid of the long tail)
• separate SNPs from indels after variant calling using SelectVariants in GATK
• accommodate possibility of false positive SNP calls around indels by using the mask option of FilterVariants with SelectVariants to remove SNPs within 6 bases of an indel