add block.LastStableOffset check ,only block.LastStableOffset >0 mean…

[xhbpiao]

When spending 0.11 kafka, block.LastStableOffset has been -1, so each time you get the data can only be parsed one. Cause serious consumer congestion

[eapache]

Is this the cause of #980?

cc @wladh

[xhbpiao]

@eapache yes

[xhbpiao]

I use filebeat to collect the log, But it's official website supports the latest version is 0.10. I have to use 0.11 protocol to parse the message, otherwise it will cause congestion。
May be the result of the reason of filebeat LastStableOffset resolution fail

[wladh]

Actually we can do away with the whole check for now. Since we're always at ReadUncommitted isolation level, the stable offset doesn't matter as we're reading everything.
Also, it's wrong as it should have been performed before returning one messages.