Record step performance

[donkirkby]

It's often helpful to know which part of a container's workflow is consuming resources, such as processing time or memory.

Proposal: after #511 provides named arguments and #920 starts recording performance, add an option for containers to configure a named argument for a current_step.csv file. Once a minute, the runcontainer command can read the last line of the file to see what the current step is, and record the time spent on each step, as well as which step triggered the peak memory usage. It might also be possible to check the memory usage of each step, but we'll have to dig into the cgroups features.