Investigate Copilot CLI server-mode timeouts in reviewer workflows

[PrzemyslawKlys]

Summary

PSTeams rollout proof shows IntelligenceX reviewer deployment is working with reviewer-20260419115614, but Copilot provider runs time out in CLI server mode on GitHub-hosted Ubuntu runners.

Evidence

• PSTeams proof PR: Verify IntelligenceX reviewer rollout PSTeams#65
• copilot_launcher=auto run: https://github.com/EvotecIT/PSTeams/actions/runs/24626672378
  • Resolves to gh copilot --
  • Times out after wait_seconds=180
• copilot_launcher=binary run: https://github.com/EvotecIT/PSTeams/actions/runs/24626773485
  • Script installer succeeds and installs /home/runner/.local/bin/copilot
  • Resolves to standalone copilot
  • Still times out after wait_seconds=180

Notes

The workflow/release deployment layer is validated: manual inputs reach the reusable workflow, the new reviewer release is downloaded, history/swarm-shadow flags are passed through, and normal OpenAI PR review remains green. The remaining problem appears to be Copilot CLI server-mode readiness/non-interactive behavior or our protocol handling around it.

Candidate Fixes

• Add stronger Copilot CLI startup/server-readiness diagnostics, including safe stderr/stdout tail capture on timeout.
• Detect auth/interactive prompts from the Copilot process and fail with a clear non-transient category instead of a generic timeout.
• Revisit whether gh copilot -- --server --stdio is a valid transport for reviewer use, or only a wrapper probe.
• Consider making Linux copilot_auto_install_method=Auto prefer script/npm when Homebrew is unavailable.