ec2tag watcher needs to back off if Client.RequestLimitExceeded is returned

If you have a reasonable number of ec2tag watchers running, you can hit the rate limiter. Rather than a fixed check_interval, there should be a way of doing some kind of backoff (as per the AWS docs). If the Client.RequestLimitExceeded is returned, then there definitely needs to be a backoff - otherwise you just continue to hammer on the door, and the rate limit exceedance will never go away.

I'm yet to discover what the rate limit is, but I hit it today with ~6 servers each watching ~24 ec2tags (plus various other odds and sods running in the background) and the default check_interval

Locate the ec2tag watcher and the handling for Client.RequestLimitExceeded, then inspect how the fixed check_interval is applied. Confirm the AWS backoff guidance and existing watcher tests before defining the behavior; done means rate-limit errors cause the watcher to back off instead of repeatedly retrying at the fixed interval.

Written by the indexing model from the issue text.