tests: Speed up test_batch_run

[rht]

This is an attempt to speed up overall CI testing time, as much as possible without reducing comprehensiveness. Currently, it takes 12s on Python 3.12 on Ubuntu.

[github-actions]

Performance benchmarks:

Model	Size	Init time [95% CI]	Run time [95% CI]
Schelling	small	🔵 +0.7% [+0.5%, +0.9%]	🔵 -0.1% [-0.2%, +0.1%]
Schelling	large	🔵 +0.0% [-1.7%, +1.9%]	🔵 +0.5% [-0.5%, +1.6%]
WolfSheep	small	🔵 +0.7% [+0.4%, +1.0%]	🔵 +0.1% [-0.0%, +0.2%]
WolfSheep	large	🔵 +0.4% [-1.2%, +2.0%]	🔵 +2.4% [+1.0%, +4.0%]
BoidFlockers	small	🔵 +0.5% [-0.1%, +1.1%]	🔵 +0.3% [-0.4%, +0.9%]
BoidFlockers	large	🔵 +0.1% [-0.3%, +0.5%]	🔵 -0.1% [-0.6%, +0.4%]

[quaquel]

Can you explain why you are concerned about CI testing time?

[rht]

1. It speeds up the review feedback loop between the reviewers and PR (that has been marked as ready) authors. The sub 1 min vs 2 min+ speedup benefit compounds for a single PR that requires several testing such as Add cell-centric discrete spaces (experimental) #1994. The Ubuntu CI runs used to be ~1 min, now about 2x faster at 30s.
2. More immediate feedback to contributors.

[quaquel]

Fair enough, although this does not make a massive difference for me. I always think of this in a more categorical manner: can I wait behind my desk, can I go for a cup of coffee, can it run over lunch, or does it take even longer? The difference between 30 seconds and 2 minutes is still very much in the behind-my-desk category.

Also, I see CI stuff as stuff that runs after having tested everything locally anyway, so I don't really see the feedback loop argument.

To be clear: speedups are still nice.

[EwoutH]

I don't see significant degradation in test functionality, so looks good. Faster CI is always nice, thanks!