BestPossibleStateCalc Stage : Fixes race condition in parallel resource computation

[ngngwr]

Issues

• My PR addresses the following Helix issues and references them in the PR description:

Fixes race condition in parallel resource computation introduced by BestPossibleState calc parallelization

Description

• Here are some details about my PR, including screenshots of any UI changes:
  What:
  Fix thread-safety issue in BestPossibleStateOutput class that caused intermittent test failures when resources are computed in parallel.
  Why:
  The _preferenceLists field in BestPossibleStateOutput used lazy initialization with a check-then-act pattern that was not thread-safe:

// BEFORE: Race condition - multiple threads could pass the null check// and overwrite each other's ConcurrentHashMap instanceif (_preferenceLists == null) { _preferenceLists = new ConcurrentHashMap<>();}_preferenceLists.put(resource, resourcePreferenceLists);

When BestPossibleStateCalcStage computes resources in parallel using StageThreadPoolHelper, multiple threads could simultaneously:

• Check _preferenceLists == null → both return true
• Create new ConcurrentHashMap instances
• One thread's map overwrites the other's, losing resource data

How:
Changed _preferenceLists from lazy initialization to eager initialization:
`

// AFTER: Thread-safe - field is initialized once at declarationprivate final Map<String, Map<String, List>> _preferenceLists = new ConcurrentHashMap<>();

`
Key changes:

• Initialize _preferenceLists eagerly at field declaration
• Make the field final to prevent reassignment
• Remove all null checks from getter/setter methods
• Simplify method implementations

Tests

No new tests added. This fix resolves existing failing tests:

TestBestPossibleStateCalcStage.testBestPossibleComputationWithMultipleResources