feat: wait for block persistence before stopping container

[skylenet]

Summary

• Poll eth_blockNumber after pre-run steps to confirm the client has finished persisting blocks before stopping/checkpointing the container. This prevents data loss with clients like reth that persist blocks asynchronously.
• Set explicit 30s graceful shutdown timeout on StopContainer for both Docker and Podman (was Docker default 10s / Podman unset).
• Add log lines around container stop/checkpoint operations for better observability.

Test plan

• Verify block persistence polling logs appear before container stop in ZFS snapshot flow
• Verify block persistence polling logs appear before checkpoint in CRIU flow
• Verify container stop uses 30s timeout (check Docker/Podman stop behavior)
• Verify graceful degradation when eth_blockNumber RPC fails (should warn and proceed)