Remove racy Darwin test.

We are going to change how the estimated latency is computed anyway,
at which point this test will no longer be valid.

Author: bzbarsky-apple

src/darwin/Framework/CHIPTests/MTRPerControllerStorageTests.m

@@ -1452,12 +1452,9 @@ - (void)doDataStoreMTRDeviceTestWithStorageDelegate:(id<MTRDeviceControllerStora
     // datastore) the device has no estimate for subscription latency.
     XCTAssertNil(device.estimatedSubscriptionLatency);
 
-    __auto_type * beforeSetDelegate = [NSDate now];
-
     [device setDelegate:delegate queue:queue];
 
     [self waitForExpectations:@[ subscriptionExpectation ] timeout:60];
-    __auto_type * afterInitialSubscription = [NSDate now];
 
     if (!disableStorageBehaviorOptimization) {
         [self waitForExpectations:@[ gotClusterDataPersisted ] timeout:60];
@@ -1504,14 +1501,6 @@ - (void)doDataStoreMTRDeviceTestWithStorageDelegate:(id<MTRDeviceControllerStora
     // some time.
     XCTAssertGreaterThan(device.estimatedSubscriptionLatency.doubleValue, 0);
 
-    // Check that this estimate is no larger than the measured latency observed
-    // above.  Unfortunately, We measure our observed latency to report end, not
-    // to the immediately following internal subscription established
-    // notification, so in fact our measured value can end up shorter than the
-    // estimated latency the device has.  Add some slop to handle that.
-    const NSTimeInterval timingSlopInSeconds = 0.5;
-    XCTAssertLessThanOrEqual(device.estimatedSubscriptionLatency.doubleValue, [afterInitialSubscription timeIntervalSinceDate:beforeSetDelegate] + timingSlopInSeconds);
-
     // Now set up new delegate for the new device and verify that once subscription reestablishes, the data version filter loaded from storage will work
     __auto_type * newDelegate = [[MTRDeviceTestDelegate alloc] init];