Handle split reads

[doubret]

Handle split reads when used with tcp input for exemple.
I replaced input buffer processing with BufferedTokenizer to handle the case when a line is not read completely.
Also, i strip the carriage return if any so that the grok can work correctly.

[elasticsearch-release]

Jenkins standing by to test this. If you aren't a maintainer, you can ignore this comment. Someone with commit access, please review this and clear it for Jenkins to run; then say 'jenkins, test it'.

[guyboertje]

@doubret - I'm sorry for not picking up this PR sooner.
This codec is going through some big changes at the moment, and the code now is not as it was when you branched.

Further, I do not believe that this codec needs this modification. Its stated goal and whole reason for being is to gather multiple lines that belong together into one event. Note the emphasis on lines - meaning that the upstream source of data really should honour this "contract" and feed this codec whole lines.
After all Its only the upstream source that "knows" about the structure and format of the data that it is receiving.