Implement AWS SDK retry and exponential backoff

[binaryblake]

Lately we've been seeing quite a few deployment errors for the AWS API rate limits being exceeded through Describe* calls in Herman. AWS recommends implementing retry and exponential backoff logic to mitigate the issue:

https://docs.aws.amazon.com/general/latest/gr/api-retries.html

Fortunately for SDK users they also have retry and backoff strategies that we can bake into the ClientConfiguration - so we shouldn't need to write our own retry logic.

@jelazos7 thanks for doing some digging on this and reaching out to AWS