tracing: revisit perf implication of many trace tags

We have an increasing amount of trace tags that we emit without any configuration. We should try to profile the cost of doing these tags, especially if there is no tracing configured (null tracer). We should potentially allow for the emitted tags to be configured in high perf scenarios.

首先定位发出 trace tag 和 null-tracer 路径的代码，然后在禁用 tracing 的情况下对 tag 处理进行性能分析。完成的标准是记录性能开销，并确定高性能场景是否需要可配置的已发出 tag。

由索引模型根据 Issue 内容生成。