fix(claw-api): defer openclaw wakes during startup

[mostlydev]

Summary

• defer openclaw-exec wakes while the target container is still in Docker health starting or unhealthy
• give OpenClaw wakes a longer exec timeout budget instead of forcing every adapter through the generic 30 second limit
• add scheduler tests for the OpenClaw timeout and health-deferral policy, and record the behavior in the changelog

Root Cause

After the cron-store repair, the remaining live scheduler failures were wake-error with detail exec timed out. claw-api was still treating every wake as a generic blocking exec with a fixed 30 second timeout, so OpenClaw startup lag and slow openclaw cron run <id> responses were being recorded as failed fires and pushed into degraded throttling.

Validation

• go test ./...

Notes

• This PR is the scheduler-side mitigation for the live Tiverton incident.
• It does not yet implement the broader container-exit/watchdog behavior for persistently unhealthy OpenClaw services.

Refs #160