feat: Let mesa runserver detect server.py as fallback

[rht]

This is part of the ongoing effort to obsolete run.py, to simplify the current model template. run.py always contains only 2 lines and hasn't been used for any customization, and so should be merged back into server.py. For discussion, see #1269.

[rht]

Tested manually:

• In the presence of both run.py and server.py
• In the presence of server.py
• In the presence of neither run.py nor server.py
  All the 3 cases works as expected.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔵 -0.2% [-0.5%, +0.1%]	🔵 -0.2% [-0.4%, -0.1%]
Schelling	large	🔵 -0.6% [-1.2%, +0.0%]	🔵 -0.4% [-1.6%, +0.7%]
WolfSheep	small	🔵 +0.2% [-0.3%, +0.6%]	🔵 +0.4% [+0.3%, +0.5%]
WolfSheep	large	🔵 +0.3% [-0.1%, +0.8%]	🔵 +0.4% [-0.3%, +1.2%]
BoidFlockers	small	🔵 +1.9% [+1.3%, +2.5%]	🔵 -0.0% [-0.6%, +0.6%]
BoidFlockers	large	🔵 +2.1% [+1.7%, +2.5%]	🔵 +0.4% [-0.0%, +0.9%]

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔴 +3.7% [+3.1%, +4.3%]	🔵 +0.6% [+0.5%, +0.8%]
Schelling	large	🔵 +6.6% [+2.3%, +11.2%]	🔵 +3.0% [-3.6%, +10.5%]
WolfSheep	small	🔵 +0.0% [-0.5%, +0.6%]	🔵 -0.3% [-0.4%, -0.2%]
WolfSheep	large	🔵 +2.0% [+1.3%, +2.9%]	🔵 +3.5% [+0.0%, +6.6%]
BoidFlockers	small	🔵 +2.8% [+1.9%, +3.6%]	🔵 -1.0% [-1.6%, -0.4%]
BoidFlockers	large	🔵 +2.6% [+2.0%, +3.3%]	🔵 -0.6% [-1.3%, +0.0%]

[rht]

@EwoutH do you know why did the performance comment happen twice?

[EwoutH]

Once for opening the PR, once for marking it "Ready for review"

[jackiekazil]

I will take a look this evening or tomorrow. This is exciting.

[EwoutH]

Can someone write a 3 sentence summary of this PR for the 2.2.5 release notes?

(I would find it really useful if each PR message is readable in isolation. Just from the PR message I would like to understand what a PR does/solves/enhances/adds without having to dig into code or other issues)

[rht]

This is a new feature, even though @tpike3 said that further older models used server.py instead of run.py, but the feature wasn't present in 2.2.4 nor 2.1.x and 2.0.x. And so can't be included in a patch release.

[rht]

But I suppose it is fine if the feature has once existed in the past? @EwoutH updated PR description. Up to you if you want to include in 2.2.5 or not.

[rht]

Wait, I just realized of a bug in this PR

In the presence of both run.py and server.py

The code says that both run.py and server.py are executed in sequence instead. Not sure why I didn't see it. Let me double check.

[rht]

I see why the code passed my manual tests: it is because the first run call blocks the execution of the next line indefinitely, and the user always presses ctrl-c to exit before the next line gets executed. All is good.