Improve Datacache observability for shared-data cluster

[wxl24life]

Feature request

A feature similar to #36865 is needed for shared-data cluster.

Is your feature request related to a problem? Please describe.

In shared data cluster, when executing show backends, the DataCacheMetrics property didn't show up metrics.

MySQL [(none)]> show backends\G
*************************** 1. row ***************************
            BackendId: 10006
                   IP: 10.0.3.3
        HeartbeatPort: 9050
               BePort: 9060
             HttpPort: 18040
             BrpcPort: 8060
        LastStartTime: 2024-02-19 21:12:14
        LastHeartbeat: 2024-02-24 11:48:01
                Alive: true
 SystemDecommissioned: false
ClusterDecommissioned: false
            TabletNum: 0
     DataUsedCapacity: 0.000 B
        AvailCapacity: 649.658 GB
        TotalCapacity: 783.211 GB
              UsedPct: 17.05 %
       MaxDiskUsedPct: 17.27 %
               ErrMsg:
              Version: 3.2.0
               Status: {"lastSuccessReportTabletsTime":"N/A"}
    DataTotalCapacity: 649.658 GB
          DataUsedPct: 0.00 %
             CpuCores: 16
    NumRunningQueries: 0
           MemUsedPct: 0.87 %
           CpuUsedPct: 0.0 %
     DataCacheMetrics: Status: Disabled
             location:
          StarletPort: 9070
             WorkerId: 1
1 row in set (0.002 sec)

Describe the solution you'd like

Describe alternatives you've considered

Additional context