Add additional peak performance benchmark metrics alongside STARS

The SRC workloads Gitlab repository hosts workloads that are representative of what users will be running at SRC sites. The STARS script collects runtimes for all of the tasks in an automated and concise way, taking an average over many runs. Runtime is only one part of understanding the impact of compute hardware on scientific pipelines. This feature looks at including metrics that track the use of the CPU, GPU, RAM, disk and power. Various software can achieve this as described below.

We need to benchmark CPUs using standard tests that measure the raw performance of single and multithreaded performance, without the overhead of additional IO or complex workloads that combine steps using single and multithreaded applications. 

We need disk IO performance measured with sustained loads outside of science tasks.

We need to track power use of systems to understand our carbon footprint if possible (this won't be a complete picture, just a start).

Weights and Biases will help track % use of single/multicore CPU, RAM/Disk use in terms of IO and capacity. This helps us understand bottle necks in tasks, and work out which tasks are dependent on specific hardware. 

• CPU benchmark software included in the STARS yaml or separate folder /bench/other
• Disk IO bench included
• Track CPU/GPU, RAM, Disk and power use inside a Weights and Biases dashboard.
• Compare to Nvidia insight or other metric tracking software (possibly not needed/too big a scope)
• Merge request and video demo