improve clarity of the client-side rate limit error when retry token bucket is exhausted

[lucix-aws]

This has caused a sizeable amount of support traffic over the years. The error message for our client-side rate limiting really reads like it's coming from the server, but it isn't.