when computing profiles at last step, need to handle profiles with no clusters

[patrickjdanaher]

This happens at the end of insitutypeML, and maybe elsewhere, causing an error.

These lines:

  profiles <- Estep(counts, 
                    clust = clust,
                    neg = neg)
  profiles <- profiles[, colnames(logliks)]

Options for handling this:

1. just don't include profiles of missing clusters (in which case they won't align to logliks, causing flightpath trouble) -- then should we output 2 loglik matrices, e.g. "logliks" and "logliks_including_lost_clusters"?
2. include the original fixedprofiles values for missing clusters -- potentially with NAs for missing genes. -- or potentially removing genes that are NA in the original fixedprofiles

... is there any reason to think we want to keep the complete logliks? Is anyone going to do anything with a lost cluster? Maybe people will want to hack with it, in which case outputting an extra logliks matrix is fine.

[patrickjdanaher]

merged into iss132-cohorting branch