Improve performance of the storage plugin under load
When Orthanc is under heavy write/upload workloads using the transfer manager, the storage plugin will occasionally become unresponsive. This appears to be because connections are not returned to the transfer manager queue. The only fix for the issue right now appears to be restarting the Orthanc server instance.
Logic should be added to the S3 transfer client so that connections timeout and go back to the queue (the timeout value should be high, maybe 180s) or that some secondary mechanism is used to ensure that the server does not lock while under load.