Seek hardware to run nightly benchmark outside of STScI

[pllim]

We currently do not have a reliable nightly benchmarking infrastructure. STScI hosting is currently broken. And even when it runs, it is on a VM beyond our control.

As discussed in Astropy Coordination Meeting 2025, possible options we could pursue:

• From US-RSE Slack (thanks to Kelle Cruz): https://blog.omsf.io/introducing-omsf-self-hosted-runners/
• Oregon State University (thanks to James Tocknell): https://osuosl.org/services/hosting/details/
  • Ethan Holz (OMSF): The OSU Open Source Lab is very cool! Also a great resource. I will note that the tool we produced requires you to bring your own cloud. More than happy to help bridge the knowledge gap if that is of interest to you!
• Ask NumFOCUS again?

Close #130

[mithr4ndir]

Hi! I have a 3-node Kubernetes home lab with dedicated resources I'd like to offer for running astropy benchmarks. I can pin a dedicated pod with reserved CPU/memory to ensure consistent, reproducible timing without noisy neighbor effects.

I'm already running nightly workloads for astronomy-commons/lsdb on the same cluster, so the operational overhead of adding ASV benchmark runs is minimal. The cluster also has a full observability stack — Prometheus, Alertmanager, Grafana, and Loki — so I can set up monitoring and alerting for benchmark failures or performance regressions.

Happy to set up the OMSF gha-runner toolkit or whatever approach works best. Is this still needed, and what are the requirements for the benchmark environment?