openclaw - 💡(How to fix) Fix Wait for compaction before failing active requester steering [2 pull requests]

Follow-up from #85489.

Active requester steering currently treats compacting from queueEmbeddedPiMessageWithOutcomeAsync as a wake failure. That means completion delivery can fall through to requester-agent/direct fallback even though the active run may become steerable again as soon as compaction finishes.

Desired behavior: compacting should not be a terminal steering failure. Wait until compaction completes, then retry steering the same wake, bounded by the existing delivery timeout/cancellation path. Other wake failures such as no_active_run, not_streaming, delivery-mode mismatch, and runtime rejection can keep their current fallback behavior.

Likely surface:

• src/agents/pi-embedded-runner/runs.ts
• src/agents/subagent-announce-delivery.ts
• src/agents/subagent-announce-delivery.test.ts

Proof expectation: regression test where an active requester is compacting, compaction ends within the delivery timeout, and generated-media completion stays on the active steering path instead of direct fallback.