[NS-66] Clean up `SparkToVectorEnginePlan` and remove dual mode caching as is currently implemented

Summary:

Clean up implementation of SparkToVectorEnginePlan
Update build.sbt to run /opt/nec/ve/bin/ps with aux to fetch info for all running VE proceses
Remove the majority of the caching code that involves dual mode column vector caches
Re-implement InVectorEngineCacheSerializer to be based on up to date infrastructure around VeColBatch creation
Remove ArrowBasedCacheSerializer as it is not in usei
Re-implement ArrowEncodingSettings as ColumnBatchEncoding and move the class to io.sparkcyclone.data package
Add tests for verifying conversions of Columnarbatch to BytePointerColVector
Add tests for verifying conversions of VeColBatch and BytePointerColBatch to BytePointerColVector

XpressAI / SparkCyclone