feat(metrics): machine_metrics_simple — CPU/RAM/disque live par machine (tâche 4)

- template machine-metrics (loadavg/nproc, /proc/meminfo, df -B1) non destructif
- parseMetrics (TDD) → cpu load/cores, mémoire kB→B + %, filesystems, warnings >=90%
- collectMetrics (SSH léger) persiste machine_metrics_latest ; getLatestMetrics (sans SSH)
- routes GET /machines/:id/metrics + POST /metrics/collect ; api latestMetrics/collectMetrics
- section Hardware : bloc métriques live (CPU/RAM/disques + alertes) + bouton Collecter
  → comble le gap « Health » de la tâche 3

tsc 0 · 108 tests · build OK.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

shared/types.ts

@@ -208,6 +208,14 @@ export interface RebootResult {
   errors?: SnapshotError[];
 }
 
+export interface MachineMetricsSimple {
+  collectedAt: string;
+  cpu: { load1: number | null; load5: number | null; cores: number | null };
+  memory: { totalBytes: number | null; usedBytes: number | null; availableBytes: number | null; usedPercent: number | null };
+  filesystems: { mount: string; fstype: string; sizeBytes: number; usedBytes: number; usedPercent: number }[];
+  warnings: string[];
+}
+
 export interface PostInstallResult {
   profilesRun: string[];
   variablesUsed: Record<string, string | number | boolean>;