Improve message pipeline functions documentation

Trying to use the `clone_message()` function, I could not find if that new message will go through the pipelines or not (it does), which is somehow confusing and may create an endless loop.

I think we should add a couple of sentences on pipeline functions creating new messages to indicate how the system will handle those new messages. We should also document how to avoid those endless loops, as they may bring the Graylog setup down (and will consume lots of CPU in any case).