Graceful Shutdown of Workers

Since September 30, 2016, there's a new method in KCL's library (for Java) that handles very well the shutdown. Its called "requestShutdown", and it has already been implemented there.

https://github.com/awslabs/amazon-kinesis-client/pull/109

We are missing that method here in Node... when KCL decides to shutdown a worker, how can we handle and perform a Graceful Shutdown? Meaning that would be cool to let the worker to finish processing the current batch, then checkpoint, and only then start the shutdown.

Start by comparing the Java KCL requestShutdown implementation in PR #109 with the Node worker lifecycle. Trace how shutdown is handled in the Node worker, then verify that the current batch finishes, a checkpoint is written, and shutdown begins only afterward.

Written by the indexing model from the issue text.