…nt to transient errors.

Bug Description

We encounter an unhandled exception which is caused by an error code 404 when callingDeleteBucketAsync to delete an existing S3 bucket. The exception is unexpected because the existing bucket should be successfully deleted and no 404 exception should be thrown.

We found that the retry mechanism of DeleteBucketAsync can actually lead to a 404 error even if the bucket exists when issuing the deletion. More concretely:

1. DeleteBucketAsync sends a HTTP request to delete the bucket
2. The bucket is deleted in the remote by the first try but a transient network errors happen (e.g., timeout), so the successful response never arrive the client, instead the transient error code 503 Server Busy may arrive Sleet.
3. The timeout (or transient error code) triggers the SDK API to retry the deletion
4. The second try fails and encounters a 404 error because the bucket has already been deleted by the first try
5. The AWS S3 SDK throws this 404 exception to Sleet

Expected behavior & Possible Solution

As implemented in this PR:

1. Ignore the 404 error when deleting the bucket.
2. Add a logging of the transient error could help improve the diagnosability.

I hope this can help! Thanks.