infra: update aggregation mode deployment

[JuArce]

infra: run aggregation mode on a daily-scheduled GPU server

Description

Run the aggregation mode on a Paperspace GPU server that is only powered on once a day, instead of keeping it running 24/7.

Type of change

• Infra

Checklist

• “Hotfix” to testnet, everything else to staging
• Linked to Github Issue
• This change depends on code or research by an external entity
  • Acknowledgements were updated to give credit
• Unit tests added
• This change requires new documentation.
  • Documentation has been added/updated.
• This change is an Optimization
  • Benchmarks added/run
• Has a known issue
  • Link to the open issue addressing it
• If your PR changes the Operator compatibility (Ex: Upgrade prover versions)
  • This PR adds compatibility for operator for both versions and do not change crates/docs/examples
  • This PR updates batcher and docs/examples to the newer version. This requires the operator are already updated to be compatible

[github-actions]

Codex Code Review

Findings

1. High Security: .github/workflows/aggregation_mode.yml stores PAPERSPACE_API_KEY as a GitHub Actions variable. This is a cloud API credential and should be a repository/environment secret: ${{ secrets.PAPERSPACE_API_KEY }}. Keep only PAPERSPACE_MACHINE_ID as a plain variable.

2. Medium Bug: infra/aggregation_mode/aggregation_mode.sh clones the repo into aligned_layer/, but the following commands still run from the parent directory. As a result, paths like aggregation_mode/proof_aggregator and ./infra/aggregation_mode/run.sh at lines 127 and 131 will not exist on a fresh setup. Fix by cloning into . or cd aligned_layer before building/copying.

3. Medium Bug: infra/aggregation_mode/run.sh does not stop on aggregation failures. If SP1 or Risc0 exits non-zero, the script still prints “finished” and reaches sudo shutdown -h now, likely making the daily run look successful while skipping aggregation. Add failure handling, for example set -euo pipefail plus a safe docker stop guard.

No significant performance issues found in this PR diff.