[FEA] Autotuner should recommend tuned settings (batchSizeBytes, concurrentGpuTasks) for OOM conditions

[mattahrens]

I would like the profiler tool recommendations to optimize for OOM scenarios where there are stats from OOM retry metrics or OOM failed tasks that should allow adjustments for Spark RAPIDS settings to reduce OOM and enhance performance.

[tgravescs]

is this on GPU ooms? Which implies the event log here was from GPU run

[mattahrens]

Yes, this was originally from GPU event logs when auto-tuner was only run with profiler on GPU runs.