JVMUtil performance improvement

https://github.com/apache/dubbo/blob/3.0/dubbo-common/src/main/java/org/apache/dubbo/common/utils/JVMUtil.java

JVMUtil. jstack方法调用整体使用OutputStream作为输出可提升性能，而内部JVMUtil.getThreadDumpString使用String长字符串拼接反而降低了性能。
建议修改内部JVMUtil.getThreadDumpString方法同样为流式处理，减少字符串拼接的性能开销

同时，getThreadDumpString方法内部，可能还考虑性能影响，增加了栈深度控制（默认32）。
这个深度控制、对性能的提升，可能没有流式处理改造能带来的性能提升高。建议应先修改流式处理以提升性能。是否可经测试评估性能影响，取消栈深度控制。以便dubbo的threaddump能够使用在各类Thread分析工具中（如flamegraph.pl等）。

Read dubbo-common/src/main/java/org/apache/dubbo/common/utils/JVMUtil.java, focusing on jstack and getThreadDumpString. Compare their output paths and measure the performance impact of stream-based handling and the current stack-depth limit. Done means the agreed performance behavior is implemented and thread-dump output remains usable by tools such as flamegraph.pl.

Written by the indexing model from the issue text.