fix: survive runtime-proxy token expiry (~1h) instead of pruning live sessions

[mikewong23571]

Problem

The runtime-proxy token expires after ~1h while the VM assignment stays alive. colab exec then gets 401/404, prints 'Session appears to be lost', deletes the local session entry — every later command says 'Session not found' although the VM is still running (visible in the Colab web UI). Re-creating sessions piles up live assignments until TooManyAssignmentsError (412).

Fix

1. State.refresh_session(): fetch fresh token/url via list_assignments() (verified: it hands out a fresh 3600s token) and update the stored session in place.
2. exec/repl/console terminal-error paths: refresh + persist credentials, exit code 2, session kept — a fresh process with refreshed credentials reconnects reliably; prune only when no live assignment matches server-side.
3. runtime.py: top-level jupyter_kernel_client.JupyterSubprotocol crashes on jkc>=1.0 (moved into wsclient); resolve defensively and omit the kwarg when unavailable.
4. Kernel-connect probe wrapped in a daemon-thread wall-clock timeout (45s) — ThreadPoolExecutor's shutdown(wait=True) still blocks on a hung connect, so a daemon thread is required.

Tested end-to-end: corrupted stored token → exec prints refresh message, keeps session, persists fresh credentials → re-run reconnects.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.