fix: cache daemon ID at init to prevent Podman token binding conflicts

Podman can return unstable or empty daemon IDs across API calls. When
the agent fetched info.ID on every report cycle, this could cause the
agent identity to change mid-session, triggering "token already in use"
errors on the server.

Cache the daemon ID at initialization and use it consistently for all
reports.

Related to #740

internal/dockeragent/agent.go

@@ -87,6 +87,7 @@ type Agent struct {
 	cfg                 Config
 	docker              *client.Client
 	daemonHost          string
+	daemonID            string // Cached at init; Podman can return unstable IDs across calls
 	runtime             RuntimeKind
 	runtimeVer          string
 	agentVersion        string
@@ -238,6 +239,7 @@ func New(cfg Config) (*Agent, error) {
 		cfg:              cfg,
 		docker:           dockerClient,
 		daemonHost:       dockerClient.DaemonHost(),
+		daemonID:         info.ID, // Cache at init for stable agent ID
 		runtime:          runtimeKind,
 		runtimeVer:       info.ServerVersion,
 		agentVersion:     agentVersion,
@@ -622,7 +624,10 @@ func (a *Agent) buildReport(ctx context.Context) (agentsdocker.Report, error) {
 
 	agentID := a.cfg.AgentID
 	if agentID == "" {
-		agentID = info.ID
+		// Use cached daemon ID from init rather than info.ID from current call.
+		// Podman can return different/empty IDs across calls, causing token
+		// binding conflicts on the server.
+		agentID = a.daemonID
 	}
 	if agentID == "" {
 		agentID = a.machineID