fix(auto-reply): retry restart input until its workspace node reconnects (#201755)

* fix(auto-reply): retry restart input until its workspace node reconnects

A restart session event that failed because the dedicated workspace node had not reconnected yet (node.invoke NOT_CONNECTED with nodeCommandDispatched false) is re-admitted every second for up to 60 seconds. Dispatched or ambiguous failures stay terminal; config re-admission keeps its own attempt budget.

* test(auto-reply): drive restart reconnect retries through pre-adoption skill discovery

* refactor(auto-reply): keep session event re-admission policy in the terminal module

The handoff module is over its line cap; the config and workspace-reconnect re-admission decision, its wait and the per-kind budgets now live next to the existing re-admission helpers. Test file formatted.

* style(auto-reply): satisfy format and lint on the restart reconnect change

* test(auto-reply): fold restart reconnect cases into the automation result suite

A separate test file added one compact CI job past the planned capacity; the cases share the automation result suite's mocked-model setup.