Match for a Throttling/Rate Limiting error

[jschniper]

I made this fork a while back and meant to submit a PR but this should help with the problem in issue #6.

[pmenhart]

Did you test this solution in high load situations?
AWS limit is 5 requests per second per log stream. We are supposed to re-try after a delay.

I implemented two alternatives in my fork:

• Remove all messages from the buffer. Transfer to AWS is delayed until the buffer fills again. Disadvantage: messages are lost.
• Explicitly delay transfer by 500msec then try again. Retrying without shedding the load brings a risk of compromising system stability.

Neither of these alternatives is bullet-proof. Therefore I am still not comfortable to offer my change as a PR for general use.

[jschniper]

My only goal with this was to prevent the hard lock in our application that we were seeing when the Elixir logger switched from asynchronous to synchronous logging. I think there is definitely more work that needs to be done but I just wanted to get this out here in case anyone else ran into the same issue.

[pmenhart]

You are right, it is definitely better than an unmatched error. We should accept your PR and add a comment/caveat to the code.
Thanks!