[FEA] Add retry to GPU timezone database caching operation

[NVnavkumar]

To address #1553 (comment)

NVIDIA/spark-rapids#9862 should cover the initial part of the comment for the database cache. But we need to make sure that there is a way to retry the allocations so we can recover from a failure when trying to call GpuTimeZoneDB.cacheDatabase().