Improve Incremental Release Benchmarks

[stanbrub]

As of now, most operational benchmarks are run in two ways; from static data, from incrementally released data. The latter has been virtually ignored in analysis of benchmarks. This is partly because incremental release doesn't compare well with other products. There is also a lack of confidence in the incremental benchmarks.

• Investigate the impact of changing the initial release (recent feedback says it's better to release too little than too much)
• Look at using a percentage of total rows, or an initial released row count based on the static run