CI: Use Python 3.13

[EwoutH]

Use Python 3.13 in all CI jobs. It’s faster and gives better error messages.

[github-actions]

Python 3.12

19:54:47 BoltzmannWealth (small) timings: Init 0.00057 s; Run 0.0672 s
19:55:09 BoltzmannWealth (large) timings: Init 0.04957 s; Run 0.6091 s
19:55:31 Schelling      (small) timings: Init 0.01169 s; Run 0.0635 s
19:55:54 Schelling      (large) timings: Init 0.07708 s; Run 0.5924 s
19:56:03 WolfSheep      (small) timings: Init 0.00565 s; Run 0.0207 s
19:56:22 WolfSheep      (large) timings: Init 0.08784 s; Run 0.3368 s
19:56:40 BoidFlockers   (small) timings: Init 0.00226 s; Run 0.2377 s
19:57:04 BoidFlockers   (large) timings: Init 0.00445 s; Run 0.7692 s

Python 3.13:

05:36:51 BoltzmannWealth (small) timings: Init 0.00058 s; Run 0.0666 s
05:37:13 BoltzmannWealth (large) timings: Init 0.04926 s; Run 0.6018 s
05:37:35 Schelling      (small) timings: Init 0.01174 s; Run 0.0628 s
05:37:59 Schelling      (large) timings: Init 0.07732 s; Run 0.6064 s
05:38:08 WolfSheep      (small) timings: Init 0.00567 s; Run 0.0208 s
05:38:26 WolfSheep      (large) timings: Init 0.08751 s; Run 0.3325 s
05:38:44 BoidFlockers   (small) timings: Init 0.00229 s; Run 0.2358 s
05:39:08 BoidFlockers   (large) timings: Init 0.00446 s; Run 0.7641 s

[EwoutH]

This is exactly why you have these kind of PRs before a new stable Python release:

• Solara fails to import with Python 3.13 on Windows widgetti/solara#797

Benchmarks numbers look remarkably similar. Where Python 3.11 and 3.12 gave a noticeable speedup, that doesn't look to be the case with 3.13.