Do not return AllNodesFailedException error to the client

[sync-by-unito]

The AllNodesFailedException is a wrapper in the driver for other others see https://github.com/apache/cassandra-java-driver/blob/4.x/core/src/main/java/com/datastax/oss/driver/api/core/AllNodesFailedException.java

We should try to understand what actual errors are being thrown in prod via splunk.

The code should return the underling error, and we should consider if there are changes to retry policy we should make because this error says “Thrown when a query failed on all the coordinators it was tried on.”

Changes to the retry policy could include changing the way the connection pool is being managed.