[ingvarch]

What changes

The client list shows a client as full as soon as an allocation is placed on it, or as free as soon as one stops. It followed the clients alone, and Nomad streams a change of a client, not of the allocations placed on it, so a full client showed on the next node event or the 30 s poll.

• The list follows Node and Allocation in every namespace. That stream needs a token that may read jobs in every namespace (event stream: fix wildcard namespace bypass hashicorp/nomad#25089), such as a management token.
• When the cluster refuses it, with an error such as a 403 or by ending it within 5 seconds with no event (how Nomad 2.0 refuses, event endpoint: use a deadline writer instead of heartbeat to close the connection on stuck writes hashicorp/nomad#27432), the list follows the clients alone, as before: no warning, and its poll timer stays as it was. Each time the list opens it asks for both again, since the token or the cluster may have changed.
• A stream of both that sent something or stayed open longer, then drops, is a drop as today: a warning and the fast poll. A refusal of the clients alone also warns and polls as today.
• The rule that tells a refused stream from one that dropped is shared with the failure note on the status line (first commit, no change in behaviour).
• docs/clients.md and docs/configuration.md say this.

Found and left for later: the screen of one client already asks for Allocation and Node in every namespace and has the same refusal with a token scoped to some namespaces (#186).

Closes #169

Checklist

• The test for this change was written first and failed
• make check is green
• The title is a conventional commit: feat(ui): ..., fix(nomad): ...