benchmarks: Add BoltzmannWealth model

[EwoutH]

BoltzmannWealth model to the benchmarks. It's probably the "best" test of Mesa that we have, since it does a lot of Mesa things and not a lot of other things.

A separate PR will update it to use AgentSet functionality, but I want to see the performance difference between that. This PR just adds the example model from mesa-examples.

---

@quaquel I noticed in that with the addition of discrete event scheduling in #2066 all models were converted to DEVS. While I understand it was useful to test, and I think it's great to have that in the benchmark for one or two models, I think for all is a bit overkill as long as it's experimental.

So the novelty is clearly in WolfSheep, I would propose keeping that on DEVS and reverting the other two back to our regular activation.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🟢 -14.8% [-15.3%, -14.4%]	🟢 -20.7% [-20.9%, -20.5%]
Schelling	large	🟢 -15.0% [-15.6%, -14.4%]	🟢 -17.8% [-18.0%, -17.5%]
WolfSheep	small	🟢 -16.1% [-17.1%, -15.0%]	🟢 -17.0% [-17.3%, -16.8%]
WolfSheep	large	🟢 -15.4% [-15.6%, -15.2%]	🟢 -16.5% [-16.7%, -16.3%]
BoidFlockers	small	🟢 -14.2% [-15.0%, -13.4%]	🟢 -9.8% [-10.5%, -9.2%]
BoidFlockers	large	🟢 -14.4% [-15.0%, -13.8%]	🟢 -12.2% [-13.0%, -11.5%]

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🟢 -13.7% [-14.2%, -13.2%]	🟢 -21.5% [-21.7%, -21.3%]
Schelling	large	🟢 -14.8% [-15.6%, -14.0%]	🟢 -11.9% [-12.7%, -10.9%]
WolfSheep	small	🟢 -13.9% [-15.0%, -12.6%]	🔵 +1.5% [-0.8%, +3.9%]
WolfSheep	large	🟢 -13.9% [-14.5%, -13.3%]	🟢 -13.6% [-15.2%, -11.8%]
BoidFlockers	small	🟢 -13.2% [-13.8%, -12.5%]	🟢 -9.4% [-10.2%, -8.6%]
BoidFlockers	large	🟢 -14.5% [-15.1%, -14.0%]	🟢 -12.9% [-13.5%, -12.4%]

[quaquel]

Good idea.

@quaquel I noticed in that with the addition of discrete event scheduling in #2066 all models were converted to DEVS. While I understand it was useful to test, and I think it's great to have that in the benchmark for one or two models, I think for all is a bit overkill as long as it's experimental.

So the novelty is clearly in WolfSheep, I would propose keeping that on DEVS and reverting the other two back to our regular activation.

Fine with me. Do we also want to have a mix of scheduler use and agentset use?

[EwoutH]

Thanks, and yes.

Can you approve? Merging is blocked until.