Handle API rate limits gracefully

[matpimenta]

This PR adds the following to the SDK:

1. Limits the number of requests to 400 per minute as per docs. The SDK waits for the next window to send more requests. It allows bursting, so it doesn't slow down requests under the limits; however, it will block when the limit is reached. It also considers the header X-Rate-Limit-Remaining to adjust the request count according to the server.
2. In case the API returns 429 Too Many Request, the SDK will retry the request using exponential backoff, with a maximum delay of 1 minute (matching the 1-minute window)

[CLAassistant]

All committers have signed the CLA.

[matpimenta]

@cla-bot check