scheduler: add ping + ssh-script check kinds, metrics refactor, 17 host check scripts

- Refactor executeCheck to return checkResult struct with metrics map
- Add ping check kind (ICMP reachability via system ping, macOS+Linux)
- Add ssh-script check kind (remote host exec via SSH, allowlisted scripts)
- Add threshold evaluation (warn/crit per metric from check config JSONB)
- Add inode tracking to disk check
- All 4 existing checks now return structured metrics
- 17 check scripts: cpu, memory, load, swap, disk_usage, disk_smart,
  updates, zfs, process, uptime, oom, journal, time, fd, docker_health,
  caddy_error_rate, backup_freshness
- Auto-deploy via tools/setup-checks.sh -> checks/install.sh on git pull
- Add ping to OpenAPI CheckKind enum and generated Go types

checks/fd_check.sh

@@ -0,0 +1,13 @@
+#!/usr/bin/env bash
+# fd_check.sh — open file descriptor usage ratio.
+set -euo pipefail
+
+FD_PCT=0
+if [ -r /proc/sys/fs/file-nr ]; then
+  read -r ALLOC _ LIMIT < /proc/sys/fs/file-nr
+  if [ "$LIMIT" -gt 0 ]; then
+    FD_PCT=$(awk "BEGIN {printf \"%.1f\", $ALLOC*100/$LIMIT}")
+  fi
+fi
+
+echo "{\"health\":\"healthy\",\"metrics\":{\"fd_pct\":$FD_PCT}}"