Add proper rate-limit and error controls to Shipbob client class

[lullis]

ShipBob's API has a rate limit of 150 requests per minute in a sliding window. The current implementation tries to account for this and also tries to respect the "Retry-after" response error in case it receives a "429 Too Many Requests" error.

The current limitation of the existing implementation:

• It retries only once in case of a 429 error.
• The client instance is the only one running with that access token. In case you have a process that instantiates many clients, the internal request counter will not be of much use.

It would be nice to improve this implementation to make it more robust, by:

• making the amount of retries configurable
• adding exponential back-off (each consecutive 429 makes the client sleep twice as long)
• adding a hard-limit to the wait time (e.g, if the client waited for a whole minute and the requests are causing 429, give up)