perf(query) Memoize the part of the logical plan tree traversal for reduced memory allocation and faster planning

[amolnayak311]

Pull Request checklist

• The commit(s) message(s) follows the contribution guidelines ?
• Tests for the changes have been added (for bug fixes / features) ?
• Docs have been added / updated (for bug fixes / features) ?

Current behavior :

Though there is no expected change in the behavior for the end users, the memory allocation for methods getPushdownShards, isTargetSchemaChanging and QueryUtils.makeAllKeyValueCombos is high. This high memory allocation can be reduced by memoizing the responses which are repeatedly called by recursing the logical plan tree

With no caching enabled the JMH tests for materialize give following results

[info] Do not assume the numbers tell you what you want them to tell.
[info] Benchmark                                        Mode  Cnt   Score    Error  Units
[info] PlannerBenchmark.benchmarkMaterializePlan        avgt    5  20.819 ± 13.145  ms/op

As seen above, getPushdownShards, isTargetSchemaChanging show up as two hotspots for memory allocation profiling.

New behavior :

After memoizing the responses, the hotspots are no longer visible. The planner's materialize is 2-3X improvement in materialize times and then two calls don't show up in allocation hotspots anymore.

[info] Do not assume the numbers tell you what you want them to tell.
[info] Benchmark                                        Mode  Cnt  Score   Error  Units
[info] PlannerBenchmark.benchmarkMaterializePlan        avgt    5  7.642 ± 3.868  ms/op