Machines on mobile: tidy host page and Nodes

Seen while checking the load meter at 448px (Task 6728): 1. Host page › Status: the StatGrid puts Nodes and OS on one line and Org alone on the next, with the org chip at a much larger font than the other values; lay the three facts out evenly and size the org chip like the Nodes table's. 2. Host page action bar at 448px: Delete wraps alone onto a second line under four buttons; keep the bar to one line on a phone (e.g. move Rename/Change org…/Delete into a ⋯ menu as the Hosts row does). 3. Hosts tab on a phone shows only the disc for state (State column gated off); put a short state word in the disc's accessible name/title or under the row when it isn't Ready, so a failed/offline host is legible without colour. 4. Nodes tab filter strip at 448px: Enabled wraps onto its own row below Search and Status, costing ~60px before the first node; fit the strip on fewer lines (e.g. a single Filters button) as other mobile lists do. Out of scope: the load meter itself (Task 6728), desktop layouts. Done means: each item rendered at 448px on the live stack, screenshotted, and judged readable; existing nodes-mobile and nodes-roster-width e2e specs green.

Phone fixes at 448px: host bar one line (Rename/Change org/Delete in ⋯), host facts one per line at one size, Hosts state word under non-Ready rows, Nodes strip = Search + Filters. Visual check passed; e2e 12/12, quick gate green. PR #5655 merged; deploy superseded. Follow-ups: draft 6775.