Worker maintenance tasks now use non-default grpcPort if configured (#8407)

Fixes #8401

When creating balance/vacuum tasks, the worker maintenance scheduler was
accidentally discarding the custom grpcPort defined on the DataNodeInfo
by using just its HTTP Address string, which defaults to +10000
during grpc dialing.

By using pb.NewServerAddressFromDataNode, the grpcPort suffix is correctly
encoded in the server address string, preventing connection refused errors
for users running volume servers with custom gRPC ports.

weed/plugin/worker/vacuum_handler.go

@@ -575,7 +575,7 @@ func buildVolumeMetrics(
 						metric := &workertypes.VolumeHealthMetrics{
 							VolumeID:         volume.Id,
 							Server:           node.Id,
-							ServerAddress:    node.Address,
+							ServerAddress:    string(pb.NewServerAddressFromDataNode(node)),
 							DiskType:         diskType,
 							DiskId:           volume.DiskId,
 							DataCenter:       dc.Id,

weed/worker/tasks/util/address.go

@@ -4,6 +4,7 @@ import (
 	"fmt"
 
 	"github.com/seaweedfs/seaweedfs/weed/admin/topology"
+	"github.com/seaweedfs/seaweedfs/weed/pb"
 )
 
 // ResolveServerAddress resolves a server ID to its network address using the active topology
@@ -16,8 +17,7 @@ func ResolveServerAddress(serverID string, activeTopology *topology.ActiveTopolo
 	if !exists {
 		return "", fmt.Errorf("server %s not found in topology", serverID)
 	}
-	if nodeInfo.Address == "" {
-		return "", fmt.Errorf("server %s has no address in topology", serverID)
-	}
-	return nodeInfo.Address, nil
+	// Note: We use NewServerAddressFromDataNode here because it encodes the GRPC port into the server address
+	// if it is non-zero, avoiding the default +10000 GRPC port offset.
+	return string(pb.NewServerAddressFromDataNode(nodeInfo)), nil
 }