Per-site per-gene estimates of selection posterior densities

'site_hpds.csv' summarizes selection coefficients obtained by renaissance counting.
It contains 12 columns:

* 'subject': Subject for the analyzed sample ('A', 'B', or 'C').
* 'sample': A sample identifier.
* 'gene': The IGHV gene and allele analyzed.
* 'variable': One of:
- 'dN' - the nonsynonymous substitution rate.
- 'dS' - the synonymous substitution rate.
- 'dNdS' - the selection coefficient, omega.
* 'subset': Subset of sequences represented by the row. One of:
- 'out-of-frame': out of frame sequences only.
- 'productive': in frame sequences without stop-codons identified.
- 'productive/out-of-frame': productive rates normalized by out-of-frame rates. See manuscript for details.
* 'site': The codon site within the V gene.
* 'imgt_site': The codon site within the V gene, using IMGT unique numbering.
* 'median': Median value
* 'hpdLower': Lower bound of the 95% HPD interval.
* 'hpdUpper': Upper bound of the 95% HPD interval.
* 'coverage': Number of sequences analyzed covering this position.
* 'oof_coverage': Number of out-of-frame sequences covering this position.

The selection coefficients presented in the manuscript are the subset of rows for which:
* 'oof_coverage > 100'
* 'subset == "productive/out-of-frame"'
* 'variable == "dNdS"'