Loading them in parallel is causing us to breach some rate limiting restrictions sporadically and the whole process fails. So this reverts to loading pages sequentially instead.
GitBook API occasionally throws a 500 when trying to get a page. This makes our whole sync Lambda fail.
This commit adds a retry step, that will wait a second and retry the GitBook API, with retries capped to three, so the Lambda is a little more resilient to random 500s from GitBook API.
