feat(node): expose bounded mission QoS and reconnect recovery

[rgthelen]

Summary

• expose mission-aware priority policy and bounded critical subscriber/fanout behavior at the PEAT node API
• authenticate critical mutations and prevent untrusted application priority from bypassing configured safety floors
• add reconnect catch-up, peer-health reset, and transitive relay controls needed by sparse multi-hop fleets
• expose priority, queue, fanout, and delivery observability used by the ARC collaboration benches
• pin the companion mesh dependency to the reviewed branch behavior

Why this is a draft

This is the node-side companion to defenseunicorns/peat-mesh#380. It was developed against an older PEAT baseline and current upstream main has since absorbed some related fixes independently. It needs a current-main rebase, dependency repin, and overlap removal before merge.

The node remains a data transport and custody service. It receives no MAVLink or aircraft-command authority.

Validation

• cargo fmt --check — pass
• 136 node unit tests — pass in the broad local run
• attachment custody, deferred delivery, multi-peer, authentication, QoS, reconnect, partition, and subscription integration tests — pass
• cargo test --test sync_subprocess_test — pass on isolated rerun (two_subprocess_sync_increments_byte_counters)
• the first broad run produced a transient localhost ConnectPeer timeout while the full mesh suite was running concurrently; the isolated exact test passed, so this PR does not hide that environmental result
• ARC local exact-four and exact-twenty mesh campaigns exercised the node behavior with routed sparse traffic and recovery

Dependency

Review this with the mesh companion first. The final rebased node branch should consume an immutable upstream peat-mesh revision rather than retain the temporary fork pin.