[KFilippopolitis]

What and why

Two JupyterHub settings for the chart. Both need the hub image from the related mip-jupyter PR; older hub images ignore them.

• Keep notebooks running across hub restarts (JUPYTERHUB_CLEANUP_SERVERS=false). The hub database already lives on jupyterhub-claim0, so a restarted hub picks running servers up again. Hub updates no longer kick users out of their notebooks or force a new login.
• Stop idle notebooks. New values jupyterhub.cullIdleTimeout (default 3600 seconds, 0 = never) and jupyterhub.cullEvery (default 600). They run the hub's idle culler. Files stay on the user's volume; open tabs, running cells and the AI chat count as activity.

How it was tested

• helm template renders both env vars; cullIdleTimeout=0 renders "0" (culler off).
• On federation-handson, with the same settings applied by hand:
  • a hub restart left the running notebook pod untouched and re-routed it;
  • the idle culler started and queries the hub API successfully.

Related PRs

• Sign in to MIP from your own computer with Keycloak mip-jupyter#8 (hub token permission and lifetime, client remote access, agent kit, hub lifecycle)
• Enforce the terms of use on the server, limit Bearer tokens to the federation's Keycloak clients, and publish the laptop sign-in client platform-backend#36 (server-side terms of use; Bearer tokens and roles limited to the federation's client)
• Account page: use MIP from your own computer platform-ui#51 (account page instructions)
• JupyterHub: keep notebooks across hub restarts, stop idle notebooks #27 (chart: keep notebooks across hub restarts, idle culler)

[ThanKarab]

LGTM!