Handle an error with a paging endpoint

[michael-richey]

What does this PR do?

If a paginated request fails try with smaller page sizes. In some cases this will solve the problem, in other cases there may be 1 resource that causes the endpoint to throw a 500 even by itself. Get as many resources as we can from the page skipping over and warning about the bad resource.

Description of the Change

Page number and page size make this very difficult to deal with. The fact that different endpoints use different methods for pagination make it worse. This approach halves the batch size until things start functioning. If batch size gets all the way down to 1 and the request still fails then skip over it. The tricky part is then to keep pulling at the smaller batch sizes until we can scale back up to the original batch size.

This change surfaced an error with team membership pagination. The PaginationConfiguration instance changes as we page through it so we can't use the same instance for each paginated request.

There's an async bug where remaining isn't always recalculated causing the loop to hang: remaining was always artificially set to 1 to get into the loop the first time. Now we immediately switch remaining to 0 until we can recalculate it. Probably need to spend some time refactoring this whole loop.

The actual github secret scanner doesn't want us to commit the cassette for the SDS because it has fake secrets in it as examples.