Nodes Health: surface /tmp inode pressure

Seen during Task 7006's visual check of the new /tmp-inodes flag (node at 92% of /tmp inodes, local stack, 2026-10-18). Suggestions: 1. The Health tab's Verdict panel and the meter's level chip still read "ok" while /tmp inodes are above 80%. Count a /tmp inode warning as a Verdict issue (e.g. "/tmp inodes 92% — mktemp will fail at 100%"). 2. The Metrics panel charts CPU, memory and disk over time, but not /tmp inode use. Add a series so an operator can see the leak rate, not just the latest figure. 3. Raise a node-health alarm (same channel as the disk alarm) when /tmp inodes cross 90%. Today the only signal is a badge someone has to notice. Out of scope: changing dispatch or placement on /tmp inode use; resizing tmpfs. Done means: - A node above 80% /tmp inode use shows the issue in the Health verdict, and the Metrics panel charts the series; `./dc check quick` green.

/tmp inodes >80% is now a Health verdict finding (chip reads warn), Metrics charts a /tmp inode series, and a node-tmp-inodes alert fires on the 90% crossing (re-arms under 80%). Quick gate green; PR #5803 merged; its deploy was superseded by a newer develop push. Follow-up draft: Task 7136.