[NancyKwongPerplexity]

Port of #837 (by @mabdelraouf-pplx) onto current main. This is the durable fix for the encrypted-suspend quarantine race, and the companion to #1760.

Change (from #837)

• Adds the node-manager CreateEncryptedSuspendSnapshot RPC. It captures an encrypted runtime and stages its ciphertext as one operation, keyed by the capture's local artifact id.
• Each step records its outcome in the staging directory. A retry after a lost reply resumes staging a confirmed capture instead of rolling back to paused, and it never captures over a capture whose outcome is unknown.
• sbx-mgr keeps the row and staging for a capture that node-manager durably owns, so recovery can complete it.

Rebase notes

• crates/libs/* → crates/l40_api/*; pplx_space_metrics::{byok,snapshot} → crate::metrics::*.
• The Cloud Hypervisor encrypted capture now passes the recorded boot source (snapshot_boot_source), because main added that parameter.
• Kept main's btrfs_send_encode_falls_back_to_full. Dropped failed_encrypted_capture_may_release_volume, as sbx-mgr: make encrypted suspend capture resumable #837 did.
• Tests now compare SandboxStatus. Cleanup-required expectations are SandboxStatus::owned(Teardown::Suspending, CleanupState::Required).

Testing

• cargo check -p pplx-space-node-mgr -p pplx-space-sbx-mgr --tests is clean.
• The node-mgr encrypt tests pass (36).
• I couldn't build the sbx-mgr test binary locally because it ran out of memory, so this relies on CI.
• Not validated on Kind/Firecracker. Left as a draft until CI and an e2e suspend/resume pass.

Supersedes #837.

_{Created with Perplexity}

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true