Log estimated phase compositions when computing driving forces with unknown phase compositions in ZPF residuals

[bocklund]

might be nice to bubble that up for debugging and plotting purposes