Add convergence visibility

[LZD-PratyushBhatt]

Compute bounded convergence reports asynchronously and expose typed cluster and resource REST
views without adding controller-path or metadata-store pressure.

Issues

• N/A. This PR is not linked to an Apache Helix GitHub issue.

Description

• This PR adds opt-in, customer-facing visibility into whether Helix assignments have
  converged.

The controller compares CurrentState with BestPossibleState and uses pending, selected, throttled,
successfully dispatched, and failed transition messages to classify each partition as:

• CONVERGED
• IN_PROGRESS
• BLOCKED
• UNKNOWN
• PAUSED

The resulting typed ConvergenceStatus records include bounded partition diagnostics, aggregate
counts, actionable reasons, WAGED target freshness, optimizer health, report age, and controller
session metadata.

Reports are persisted under:

/{cluster}/PROPERTYSTORE/HELIX_CONVERGENCE_STATUS
/{cluster}/PROPERTYSTORE/HELIX_CONVERGENCE_STATUS/{resource}

The following REST endpoints expose the reports:

GET /clusters/{cluster}/convergence
GET /clusters/{cluster}/resources/{resource}/convergence

Performance and safety properties:

• Monitoring is disabled by default through CONVERGENCE_MONITORING_ENABLED.
• Full partition calculation and persistence run on a dedicated asynchronous worker.
• Pending controller events are deduplicated.
• Controller pipeline submission measured 0 ms in normal live testing, occasionally 1 ms.
• Persistence is rate-limited to at most once per second.
• Unchanged reports are not rewritten.
• Existing convergence records are loaded once per controller session and cached.
• Diagnostic output is bounded to 20 partitions and 64 assignment entries per side.
• A 10,000-partition bounded calculation completed in approximately 54 ms off-thread.
• Writes are controller-session fenced and use an incomplete-root/complete-root protocol.
• REST rejects missing, incomplete, unsupported, or stale-controller reports.

There are no UI changes.

Tests

• The following unit and integration tests cover this change:

• TestConvergenceStatusCalculator

• TestConvergenceStatusPersistStage

• TestWagedRebalanceStatus

• TestConvergenceStatusResponseMapper

• TestConvergenceStatusAccessor

• TestPropertyPathBuilder

• TestPropertyKeyGetPath

Core focused test result:

mvn -pl helix-core \
  -Dtest='TestPropertyPathBuilder,TestConvergenceStatusPersistStage,TestConvergenceStatusCalculator' \
  test

Tests run: 13, Failures: 0, Errors: 0, Skipped: 0
BUILD SUCCESS

REST end-to-end test result:

mvn -pl helix-rest -am \
  -Dtest='TestConvergenceStatusAccessor,TestConvergenceStatusResponseMapper' \
  -Dsurefire.failIfNoSpecifiedTests=false test

Tests run: 4, Failures: 0, Errors: 0, Skipped: 0
BUILD SUCCESS

Repository build result:

mvn clean install -Dmaven.test.skip.exec=true

BUILD SUCCESS

Live validation was also performed against a standalone ZooKeeper server with real controller,
participant, and REST processes. The scenarios covered:

• Initial assignment convergence
• Dynamic resource addition
• Participant scale-out, loss, and rejoin
• Cluster pause and resume
• Maintenance mode entry and exit
• Controller loss and recovery with stale-report detection
• REST restart using persisted reports
• WAGED enablement and instance movement
• Resource deletion and PropertyStore cleanup
• A deliberately long-running participant transition that remains IN_PROGRESS

Changes that Break Backward Compatibility (Optional)

• None.

All changes are additive. Convergence monitoring is disabled by default, existing clusters do not
perform convergence calculation or persistence unless explicitly enabled, and existing REST APIs
are unchanged.

Documentation (Optional)

• N/A. No wiki page was added.

The REST contract and configuration behavior are documented by the model, endpoint implementation,
and test coverage in this PR.

Commits

• The commit follows the project commit-message guidelines:
  1. Subject is separated from the body by a blank line.
  2. Subject is under 50 characters.
  3. Subject does not end with a period.
  4. Subject uses the imperative mood.
  5. Body explains what and why.

No issue reference is included because this PR is not linked to an Apache Helix issue.

Code Quality

• The diff follows existing Helix Java formatting conventions.
• git diff --check passes.
• The code compiles for the repository's JDK 8 and JDK 11 build targets.