What is the problem the feature request solves?

When I profiled tpcds q64, I found that update_metrics took 8.9 % of task time.

Describe the potential solution

I propose the following improvements to optimize update metrics:

• Define a NativeMetricNode proto type to pass all metric nodes at once to avoid iterative jni calls.
• Call update metrics when releasing plan to reduce the number of calls.

Additional context

No response