HdfsReader的线程数量，通过通道数指定无效，是根据文件数量决定的。

文档中写到"HdfsReader支持使用多线程进行数据抽取。线程并发数通过通道数指定。"


但是看代码，是根据文件数量决定的。(后续引起了TxtFileWriter生成了多个文件)

Start at the HdfsReader implementation and compare its thread-count logic with the documentation statement quoted in the issue. Reproduce the behavior with different file and channel counts, then make the channel-count contract consistent and verify that TxtFileWriter no longer receives unintended parallel output.

Written by the indexing model from the issue text.