Fix health tab: node identity indicator + cluster card overflow

Add node_id to metrics snapshot so the health tab shows which instance
the user is viewing from. Hoist nodeID computation so it works on both
SQLite and Postgres deployments. Fix cluster node card overflow with
text-overflow ellipsis and proper flex constraints.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

server/metrics/collector.go

@@ -18,6 +18,7 @@ import (
 
 // Snapshot is the top-level JSON response from GET /api/v1/admin/metrics.
 type Snapshot struct {
+	NodeID     string            `json:"node_id"`
 	Runtime    RuntimeMetrics    `json:"runtime"`
 	DB         DBMetrics         `json:"db"`
 	Cluster    *ClusterMetrics   `json:"cluster,omitempty"`
@@ -91,6 +92,7 @@ type TriggerFireCountFunc func() int64
 
 // Collector gathers metrics on demand for the admin endpoint.
 type Collector struct {
+	nodeID           string
 	db               *sql.DB
 	hub              ConnCounter
 	bus              BusCounter
@@ -101,8 +103,9 @@ type Collector struct {
 }
 
 // NewCollector creates a metrics collector with all required dependencies.
-func NewCollector(db *sql.DB, hub ConnCounter, bus BusCounter, stores store.Stores, sandboxFn SandboxStatsFunc, triggerFn TriggerFireCountFunc, startTime time.Time) *Collector {
+func NewCollector(nodeID string, db *sql.DB, hub ConnCounter, bus BusCounter, stores store.Stores, sandboxFn SandboxStatsFunc, triggerFn TriggerFireCountFunc, startTime time.Time) *Collector {
 	return &Collector{
+		nodeID:           nodeID,
 		db:               db,
 		hub:              hub,
 		bus:              bus,
@@ -115,7 +118,7 @@ func NewCollector(db *sql.DB, hub ConnCounter, bus BusCounter, stores store.Stor
 
 // Collect gathers all metrics synchronously and returns a snapshot.
 func (c *Collector) Collect(ctx context.Context) *Snapshot {
-	snap := &Snapshot{}
+	snap := &Snapshot{NodeID: c.nodeID}
 	snap.Runtime = c.collectRuntime()
 	snap.DB = c.collectDB(ctx)
 	snap.Cluster = c.collectCluster(ctx)