flume-hdfs-sink Transmission speed

For each event, a task will be submitted to the callTimeoutPool thread pool during append, and then the future.get method will be used to obtain the result. This may cause a large load on the thread pool, which cannot be processed at all. Through flame graph analysis, it can also be seen that retrieving tasks from the thread pool takes a lot of time. Can we change it to synchronous

Start in the flume-hdfs-sink append path, focusing on callTimeoutPool task submission and the subsequent future.get call. Review the flame graph and current behavior, then determine whether synchronous handling is appropriate; done should address the reported thread-pool load and transmission-speed concern, with validation based on the observed performance.

Written by the indexing model from the issue text.