Add randomized delay between attempts to ExponentialReconnectionPolicy

We have a cluster running with multiple nodes and there are around 10 to 60 different client instances connecting to these nodes depending on the load with a connection pool of core size 8. When one of the Cassandra nodes is unresponsive for a time larger than the connection timeout, the client instances disconnect and attempt to re-create the connections. When the number of client instances running is high, we think it might be creating high load on the node because re-connection attempts are synchronized (meaning all instances are trying to re-connect at once during different attempts).

I was checking the ExponentialReconnectionPolicy in 4.x and ExponentialReconnectionPolicy in 3.x and these don't include jitter. What do you think about adding such capability to the driver?