Handle missing SuperPMI replay MCH inputs as a non-failing skip in CI

[Copilot]

SuperPMI replay shards for windows/arm64 started failing when a new JIT/EE GUID had no MCH payloads in Azure Storage. The replay path treated “no MCH available” as a hard error and failed affected partitions.

• What changed

  • Added a replay-only switch in superpmi.py: --skip_if_no_mch.
  • In replay mode, when MCH resolution returns empty and this switch is set, replay now logs a warning and exits success instead of returning failure.
  • Updated superpmi_replay.py to pass --skip_if_no_mch when invoking superpmi.py replay for Helix replay partitions.

• Scope

  • Behavior change is intentionally limited to replay mode (the CI path in this issue).
  • Existing default behavior remains unchanged when the switch is not provided.

• Key code path

  # superpmi.py (replay mode)
  mch_files = get_mch_files_for_replay(local_mch_paths, coreclr_args.filter)
  if mch_files is None:
      if coreclr_args.skip_if_no_mch:
          logging.warning("No MCH files found for replay; skipping.")
          return 0
      return 1

[dotnet-policy-service]

Tagging subscribers to this area: @dotnet/runtime-infrastructure
See info in area-owners.md if you want to be subscribed.