fix(runtime): preserve the workbench across reconnects

[luoye520ww]

Summary

• surface an explicit degraded state while health checks are failing without treating it as a crash loop
• keep the workbench, thread history, and composer draft mounted through restarting/offline/failed runtime states
• add a reconnect action for terminal runtime connection states
• coalesce concurrent active-turn recovery requests so a reconnect cannot replay the same turn twice
• only consume composer text and attachments after the runtime accepts the turn

Validation

• npm.cmd exec -- vitest run src/main/kun-runtime-supervisor.test.ts src/renderer/src/components/RuntimeStatusBanner.test.ts src/renderer/src/lib/runtime-banner-visibility.test.ts src/renderer/src/store/chat-store-thread-actions.test.ts src/renderer/src/components/workbench/useWorkbenchComposerSubmitController.test.ts
• npm.cmd run lint -- --no-warn-ignored
• npm.cmd run typecheck
• npm.cmd run build

The Kun package typecheck was also attempted; this checkout's borrowed dependency tree is missing existing packages (@xmldom/xmldom, yazl, etc.), while the targeted Kun tests pass.

[luoye520ww]

Superseded by consolidated runtime recovery PR #947, which passed Typecheck, Linux, macOS, and Windows packaged checks. Closing this duplicate to keep the review surface focused.