Allow reconnect after settled initialization uncertainty

[SaladDay]

When Runtime initialization returned unknown, ordinary disconnection kept retrying an already closed Router forever even after its apply and receipt workers had joined. Shutdown now permits reconnect after these workers finish, while preserving the unknown result and the old Router's admission fence. Core's persisted initialization failure still prevents setup replay and execution; workspace-write and native-executor cleanup keep their existing settlement checks.

Validation: dispatch, CLI, localworkspace and clirunner packages; preparation race tests repeated three times; affected vet; real PostgreSQL initialization cases covering reconnect after Worker restart without replay; English/Chinese translation checks. Shutdown regressions were red before the fix and green after it, including blocked apply and receipt delivery. Fresh independent review found no blocking issue. The native process checks cover the existing process-group/Job owner, not arbitrary escaped descendants or exactly-once external effects.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}