Reject some % of requests with acme:error:rateLimited

[sophie-h]

Preferably with a 'Retry-After' header

[shred]

I'd like that! However, to ensure that integration tests against Pebble are deterministic, there should always be a way to eventually get a good response in a timely manner (this is, within a few seconds). Maybe the rateLimited error ratio and retry delay should be controlable via the Management Interface (issue #74)?

[jvanasco]

This feature is especially relevant given recent LetsEncrypt changes to throttle at certain hours.

Perhaps the presence of a specific header could bypass this, so client developers could set this feature to 100% to ensure they are compatible.

i.e. my client would catch this and "sleep()" in production, but in staging I would catch this and set "x-retry-ignore=true" or something like that.