Scaling some processes based on sample #

[cjfields]

Several initial steps in the workflow need reasonable resource scaling based on the input sample number. A recent run with 1400 samples from a large NovaSeq run have run into both memory and timing issues, which required bumping up the requirements in the config file. We could possibly benchmark subsets of these to get some run metrics.