Decoder might block if there is data remaining to be flushed from the codec and the stream blocks

[Nemo157]

See gyscos/zstd-rs#119, I'm pretty sure the local bufread::Decoder implementations suffer from this same issue. My first attempt at writing a test case failed to fail though.

[nl5887]

I encountered this issue when keeping an open connection, while not sending data. This caused the remaining data to be kept in the decoder. I have fixed the issue, but will try to write a test case.

[fr3akX]

could you, please, review and merge this PR?

[link2xt]

Seems tokio implementation was not fixed, started to fix it in #294 in a different way and then discovered this change.