perf(index): Use Set instead of ArrayList to reduce memory overhead in key lookup

[prashantwason]

Describe the issue this Pull Request addresses

This PR optimizes the bloom index key lookup by using Set instead of ArrayList for storing candidate record keys. ArrayList has large memory overhead which occurs when the ArrayList grows beyond its initially allocated size. Set is better suited to an exists check and avoids the need to copy the collection when calling filterRowKeys().

Summary and Changelog

• Changed candidateRecordKeys in HoodieKeyLookupHandle from ArrayList<String> to HashSet<String>
• Updated filterKeysFromFile method signature in HoodieIndexUtils to accept Set<String> instead of List<String>
• Removed unnecessary .stream().collect(Collectors.toSet()) call since the input is already a Set

Impact

No public API changes. This is an internal optimization that reduces memory overhead when looking up a large number of keys during bloom index operations.

Risk Level

low - This is a straightforward type change from ArrayList to HashSet with no behavioral changes. The Set semantics are actually more appropriate since we're checking for key existence.

Documentation Update

none

Contributor's checklist

• Read through contributor's guide
• Enough context is provided in the sections above
• Adequate tests were added if applicable

[hudi-bot]

CI report:

• 5e4ce50 Azure: FAILURE

Bot commands

@hudi-bot supports the following commands:

• @hudi-bot run azure re-run the last Azure build