ci: non-GPU instances

[redpanda-f]

CI job is transitioning away from 16xlarge+gpu to non-GPU instances.

Test Type:

• Basic E2E foc-devnet setup with 1 curio SP (integration test)

Trials:

• Top level GH Actions run: https://github.com/FilOzone/foc-devnet/actions/runs/21510443918
  -8xlarge+amd means c5a.8xlarge 32 vCPU, 64 GB RAM, AMD EPYC 7R32 @ 3.3 GHz
  • Verdict: ❌ FAIL
  • GH Actions Run: https://github.com/FilOzone/foc-devnet/actions/runs/21510443918/job/61975929712
    -9xlarge means c5.9xlarge 36 vCPU, 72 GB RAM, Intel Xeon Platinum 8124M @ 3.4 GHz
  • Verdict: ❌ FAIL
  • GH Actions Run: https://github.com/FilOzone/foc-devnet/actions/runs/21510443918/job/61975929707
    -12xlarge+amd means c5a.12xlarge 48 vCPU, 96 GB RAM, AMD EPYC 7R32 @ 3.3 GHz
  • Verdict: ❌ FAIL
  • GH Actions Run: https://github.com/FilOzone/foc-devnet/actions/runs/21510443918/job/61975929709

[Copilot]

Pull request overview

This PR updates the CI configuration to run integration tests on a different class of self-hosted runner, moving away from GPU-specific instances to CPU-focused ones.

Changes:

• Adjusts the integration-test job in .github/workflows/ci.yml to use the ["self-hosted", "linux", "x64", "12xlarge+amd"] runner label instead of ["self-hosted", "linux", "x64", "16xlarge+gpu"].