[Bug] delegation task failure triggers excessive retrigger messages

Description

When Agent A delegates a task to Agent B (via send_agent_message tool), if B fails to complete the task, the system automatically injects a delegation completion retrigger to wake up A for handling the failure.

Current implementation has issues:

1. No cooldown period: retrigger fires continuously at ~20-30 second intervals
2. No retry limit: keeps retrying indefinitely after failure
3. Every retrigger sends message: even when A tries not to reply, the retrigger result gets routed to the user's channel

Impact

Users may receive a large number of duplicate messages (e.g., 1000+) in a short period, severely degrading the experience.

Expected behavior

1. On delegation failure, A should only report failure to user once
2. Retrigger should have cooldown and retry limits
3. Prevent message storms caused by mechanism flaws

Environment

• spacebot v0.3.1

Start by tracing the send_agent_message path and the delegation completion retrigger handling described in the issue, then inspect how failed delegations are routed to the user channel. Done means failures report to the user once, retriggers have cooldown and retry limits, and repeated failure does not create a message storm.