Should we clean up the benchmark history?

[kandersolar]

The benchmark results history is getting a bit confusing. The plots are supposed to quickly draw the eye to performance jumps, but the majority of jumps are just due to environment changes. Example: SolarPositionNumba.time_sun_rise_set_transit_spa

Although changes in performance due to environment is surely interesting, I think the main goal for our benchmarks should be for evaluating performance changes in pvlib itself (holding the environment ~constant). Should we wipe out and rebuild the history using the current set of environments?

We may also want to change our environment management strategy to reduce future fragmentation.

[cwhanse]

Hard to decipher with that many traces. Do we need the 3 time options (1, 10, 100 days)? I agree that the main purpose is to track changes in pvlib code performance.