Harvest should describe how metrics are calculated and show comparision with CLI, KeyPerf, and PAS

[cgrinds]

This wiki link compares CLI and Harvest, but the screenshots and text are out-of-date. Instead of fixing that link,

• create a new page on the public docs site that describes how a latency counter is cooked
• explain how a certain number of IOPs are required to make sense of the latency and the special cases for that threshold
  • explain how this thresholding applies to any counter that has a property of average or percent and has latency in the name
• include screenshots of Prometheus (or Grafana), ONTAP CLI statistics, PAS, and System Manager screenshots for the same time period
• Delete wiki page and update references to that page to point to above public documentation