Makes HttpLogger flushable and/or sync optionally

[jcchavezs]

From @nathanhoel at https://gitter.im/openzipkin/zipkin?at=5edbbf82a91f120a6ce91405

Hi, I have read #362
I have the same use case as the person who posted this issue.
I see that you added flush() to BatchRecorder - but as far as I can tell this only sends the messages to the logger such as HttpLogger, which then queues the messages and then asynchronously sends them to the collector.
Since processQueue() on HttpLogger is non-blocking I am still in the same situation, where my AWS Lambda will freeze the execution environment before the messages are ever sent to the collector.
I feel like I may be not understanding something because you added the flush() method specifically to handle this situation, but from my perspective it seems like the problem still exists.
Any advice would be much appreciated, thanks!

[jcchavezs]

So the option is then to either make the httplogger to be sync when passing httpInterval=0 or add a flush method to httplogger. Either change works for your use case but they both can be implemented to cover some other use cases. Would you be up to open a PR for the change? In any case if you jump in the issue describing this very problem I can reopen it and we can discuss the solution there.

[jcchavezs]

I think a flush method would be ideal.