Hikari init additional endpoints

[Anay-Mahajan]

Summary

• TPC-C 1k chaos jobs failed at --execute startup with HikariPool$PoolInitializationException (NoRouteToHost / Connection refused). Chaos often stops the first IP in --nodes before Hikari finishes pool init. The Yugabyte smart driver only learns topology after a connection succeeds, so that first pool had nowhere else to try and the process exited before any transaction ran.
• For Yugabyte, each Hikari pool now gets the other --nodes as dataSource.additionalEndpoints so the first connect can land on a live tserver. initializationFailTimeout=0 lets the pool start if nothing answers yet; a connection that opens but fails validation (for example bad credentials) still fails fast.
• This does not replace smart-driver load balancing. Extra endpoints are only used to establish the first connection. After that, routing still comes from yb_servers().
• Verified on PerfService for node, zone, and region chaos. Execute got past Hikari init in all three:
  • Node
  • Zone
  • Region