May underestimate requests (unreliable use of SCAN)

[asmello]

In respectful_requester.py#L189, the pattern len(redis.scan()) is used.

According to the redis-py library documentation, scan() does no iteration itself, you are expected to use the returned cursor to fetch the remaining results manually. Line 189 seems to assume that passing a high enough count argument will ensure all results are returned, but according to the Redis documentation that is not the case.

The consequence is that _requests_in_timespan() might underestimate the number of requests in the timespan, making rate control unreliable.

A simple fix is to replace the scan() call with scan_iter(), which does ensure that the full result set is returned.