Call to refresh_connections only if failed_conn isn't empty

[barshaul]

ATM refresh connections is being called from each periodic check even when no failed connections were found which resulting in a lot of logs like:

�[2m2024-07-03T17:07:12.769490Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m Started refreshing connections to []
�[2m2024-07-03T17:07:12.769523Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m refresh connections completed
�[2m2024-07-03T17:07:14.331242Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m Started refreshing connections to []
�[2m2024-07-03T17:07:14.331271Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m refresh connections completed
�[2m2024-07-03T17:08:12.771693Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m Started refreshing connections to []
�[2m2024-07-03T17:08:12.771726Z�[0m �[32m INFO�[0m �[2mredis::cluster_async�[0m�[2m:�[0m refresh connections completed

[avifenesh]

LGTM