fix: set _waitForDelete=false in moveChunk to prevent ExceededTimeLimit

[hyeonghwa]

Problem

During performance testing on the router (mongos), the following error occurred repeatedly:

Command failed with error 96 (OperationFailed):
'Data transfer error: ExceededTimeLimit: Failed to delete orphaned POCDB.POCCOLL
range [{ _id: { w: 1, i: 1 } }, { _id: MaxKey }) :: caused by :: operation exceeded time limit'

Root Cause

In MongoWorker.java, the moveChunk command was issued with _waitForDelete: true:

admindb.runCommand(new Document("moveChunk", ...)
    .append("_secondaryThrottle", true)
    .append("_waitForDelete", true)   // <-- root cause
    .append("writeConcern", new Document("w", "majority")));

_waitForDelete: true forces moveChunk to block synchronously until all orphaned documents are deleted from the donor shard after chunk migration.

When the orphan range is large — in this case [{ _id: { w: 1, i: 1 } }, { _id: MaxKey }), which spans to the last chunk — the deletion operation exceeds the server's operationTime limit, resulting in error code 96 (OperationFailed: ExceededTimeLimit).

Fix

Changed _waitForDelete from true to false so that orphan cleanup runs asynchronously in the background, allowing moveChunk to return immediately without blocking on deletion.

// Before
.append("_waitForDelete", true)

// After
.append("_waitForDelete", false)

Impact

• moveChunk no longer times out during performance tests on sharded clusters with large orphan ranges.
• Orphaned documents are still cleaned up by the MongoDB balancer in the background — no data loss occurs.
• This is consistent with MongoDB's default behavior when _waitForDelete is omitted.

Test Plan

• Verified fix resolves ExceededTimeLimit error on router during POCDriver performance test
• Run full performance test suite (Insert / Key Query / Range Query / Update) on sharded cluster