Automated Benchmarking

In anticipation of regular benchmarking runs we should design a system for automated Benchmarking.

This includes:

* A process to orchestrate the benchmark (provision necessary cluster resources, generate load, manage results gathering, persist results)
* A format in which results will be persisted (should be flexible enough to accommodate adding/removing metrics as product understanding grows)

This system should run as a simple shell command and output the metrics results to files in order to streamline integration with most CI/CD systems.

ClusterLoader2 is a good example of the general idea of the I/O and execution model, though CL2 has a much broader scope than this proposal intends.