Full join instead of inner join cohorts together

Current code inner joins SNP-Gene pairs recursively across cohorts, meaning that the eventual meta-analysis can only be run on the SNP-Genes present in every cohort. This was due to RE2C failing on missing values. 

ACTION:

Swap out inner join to a full join, figure out how to make RE2C happy. If necessary, let RE2C step throw out SNP-Genes with missing rows. Not ideal.