optimize reduce CPU usage for multi-thread decompressor

[sisong]

ref ebiggers#335 (comment) ， a speed testing
The multi-thread parallel of pgzip mainly optimizes the compression speed.
and the decompression multi-thread, only handles asynchronous I/O, and it supports up to 3 threads: one for reading the code data, one for decompressing, and one for writing the uncompresssed data.
For the sake of source code simplicity, the waiting between threads is loop+this_thread::yield(), and the formal implementation should use condition variable.

[sisong]

@ghuls
update for reduce CPU usage when multi-thread decompress.