Fix flaky ClusterMigrateRangeIndexCheckpointThenRecover: flush AOF before graceful restart

[tiagonapoli]

Problem

ClusterMigrateRangeIndexCheckpointThenRecover (in Garnet.test.cluster.migrate.rangeindex) intermittently fails with "post-checkpoint RI.SET should survive recovery".

Root cause

The test does a post-checkpoint RI.SET and then immediately restarts the target node. That write is acknowledged to the client but, at CommitFrequencyMs=0 (async auto-commit), is not yet durably flushed to the AOF. A graceful shutdown (GarnetServer.Dispose → StoreWrapper.Dispose) cancels the background committer and does not flush the AOF, so the acked-but-uncommitted tail is lost on recovery — the test then can't find the post_ckpt value.

This mirrors the durability contract every other AOF-recovery test relies on (they all issue COMMITAOF / commitWait / CommitAOFAsync before restart); this test omitted it.

Fix (test-only)

Add an opt-in ensureAofFlush parameter to ClusterTestContext.ShutdownNode / RestartNode that issues a synchronous Store.CommitAOFAsync before disposing the node. The RangeIndex migrate test's RestartNode helper passes ensureAofFlush: true so all its restarts flush pending AOF commits first.

Validation

• Builds clean (0 warnings).
• Previously-flaky test passed 7/7 locally.
• Docker CPU-throttled repro fleet run to confirm the intermittent failure no longer reproduces (in progress).

[Copilot]

Pull request overview

This PR addresses intermittent failure in the cluster RangeIndex migration recovery test by ensuring pending AOF work is durably committed before a “graceful” node restart during the test.

Changes:

• Added an opt-in ensureAofFlush parameter to ClusterTestContext.ShutdownNode(...) and RestartNode(...) to force an AOF commit prior to disposal.
• Updated the RangeIndex migrate test helper to restart nodes with ensureAofFlush: true.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
test/cluster/Garnet.test.cluster/ClusterTestContext.cs	Adds optional AOF commit before shutdown/restart to make test restarts durable.
test/cluster/Garnet.test.cluster.migrate.rangeindex/ClusterRangeIndexMigrateTests.cs	Uses the new option to flush AOF before restarting the target node in the flaky test.