Fill in missing values after reading input data

This is a minor revision to the underlying code which does one thing -- automatically fills in missing values for the input data after normalization.

I noticed these missing values occurring in just 3 cells of a matrix with ~400 samples and ~4.5M genes, and they only appeared when normalizing by the sum of the depth of each sample. Ideally I would have tracked down exactly why these values appeared, but I could not replicate the error in testing data of any reasonably small size.

In this release, the logs record how many cells contain missing values which are filled in, and this does not stop the overall analysis, as it did previously.

Another minor revision to the Docker image is that the find-cags code is now installed from the repo, which just helps with testing and evaluation within the Docker image.