We should consider adding support for sampling request logs, in the style of: https://github.com/jeremydaly/lambda-api#sampling