Incorrect msgBacklog value

**Describe the bug**
Despite there are no messages in the backlog, positive value was set to "msgBacklog" in the response of partitioned-stats(API).

I think "msgBacklog" is calculated from "ManagedLedgerImpl.ENTRIES_ADDED_COUNTER_UPDATER" and "ManagedCursorImpl.messagesConsumedCounter".
https://github.com/apache/pulsar/blob/dc7d01efc6cf2df5631bc509758f2212bede35ce/managed-ledger/src/main/java/org/apache/bookkeeper/mledger/impl/ManagedCursorImpl.java#L736

Looking at the debug log, [subscriptionName1] appears to have the wrong "ManagedCursorImpl.messagesConsumedCounter" value.
In contrast, [subscriptionName2] looks correct.

Log:
```
12:08:44.575 [prometheus-stats-33-1] DEBUG o.a.b.mledger.impl.ManagedCursorImpl - [] Consumer cursor ml-entries: 554102 -- deleted-counter: 549461 other counters: mdPos 27058740:10 rdPos 27058740:11
12:08:44.575 [prometheus-stats-33-1] DEBUG o.a.b.mledger.impl.ManagedCursorImpl - [] Consumer cursor ml-entries: 554102 -- deleted-counter: 554102 other counters: mdPos 27058740:10 rdPos 27058740:11
```

Stats:
```
{
"msgRateIn" : 3.283531005730665,
"msgThroughputIn" : 1657.249768369514,
"msgRateOut" : 3.2835310293187825,
"msgThroughputOut" : 1657.2497802748082,
"averageMsgSize" : 504.7157360406091,
"storageSize" : 142581,
"publishers" : [ ],
"subscriptions" : {
"" : {
"msgRateOut" : 3.2835310293187825,
"msgThroughputOut" : 1657.2497802748082,
"msgRateRedeliver" : 0.0,
"msgBacklog" : 4677,
"blockedSubscriptionOnUnackedMsgs" : false,
"unackedMessages" : 0,
"type" : "Exclusive",
"activeConsumerName" : "4efd5",
"msgRateExpired" : 0.0,
"consumers" : [ {
"msgRateOut" : 3.2835310293187825,
"msgThroughputOut" : 1657.2497802748082,
"msgRateRedeliver" : 0.0,
"consumerName" : "4efd5",
"availablePermits" : 946,
"unackedMessages" : 0,
"blockedConsumerOnUnackedMsgs" : false,
"metadata" : { },
"connectedSince" : "2019-09-02T13:29:21.891+09:00",
"clientVersion" : "2.1.1-incubating",
"address" : "/:55920"
} ]
}
},
"replication" : {
"" : {
"msgRateIn" : 3.283531005730665,
"msgThroughputIn" : 1657.249768369514,
"msgRateOut" : 0.0,
"msgThroughputOut" : 0.0,
"msgRateExpired" : 0.0,
"replicationBacklog" : 0,
"connected" : true,
"replicationDelayInSeconds" : 0,
"inboundConnection" : "/:36792",
"inboundConnectedSince" : "2019-08-29T11:36:48.612+09:00",
"outboundConnection" : "[id: 0xb6500383, L:/:50306 - R:/:6651]",
"outboundConnectedSince" : "2019-08-29T11:36:48.64+09:00"
}
},
"deduplicationStatus" : "Disabled"
}
```

Stats-internal:
```
{
"entriesAddedCounter" : 552514,
"numberOfEntries" : 611632,
"totalSize" : 1007353769,
"currentLedgerEntries" : 15460,
"currentLedgerSize" : 57848652,
"lastLedgerCreatedTimestamp" : "2019-09-03T08:04:37.87+09:00",
"waitingCursorsCount" : 2,
"pendingAddEntriesCount" : 0,
"lastConfirmedEntry" : "27010835:15459",
"state" : "LedgerOpened",
"ledgers" : [ {
"ledgerId" : 25679299,
"entries" : 2544,
"size" : 11803597,
"offloaded" : false
}, {
"ledgerId" : 25711440,
"entries" : 11923,
"size" : 42865506,
"offloaded" : false
...(Omitted...)
}, {
"ledgerId" : 26992239,
"entries" : 143,
"size" : 658467,
"offloaded" : false
}, {
"ledgerId" : 27010835,
"entries" : 0,
"size" : 0,
"offloaded" : false
} ],
"cursors" : {
"" : {
"markDeletePosition" : "27010835:15409",
"readPosition" : "27010835:15460",
"waitingReadOp" : true,
"pendingReadOps" : 0,
"messagesConsumedCounter" : 547823,
"cursorLedger" : 27014015,
"cursorLedgerLastEntry" : 426,
"individuallyDeletedMessages" : "[]",
"lastLedgerSwitchTimestamp" : "2019-09-03T09:34:14.211+09:00",
"state" : "Open",
"numberOfEntriesSinceFirstNotAckedMessage" : 51,
"totalNonContiguousDeletedMessagesRange" : 0,
"properties" : { }
},
"" : {
"markDeletePosition" : "27010835:15459",
"readPosition" : "27010835:15460",
"waitingReadOp" : true,
"pendingReadOps" : 0,
"messagesConsumedCounter" : 552514,
"cursorLedger" : 26992238,
"cursorLedgerLastEntry" : 848,
"individuallyDeletedMessages" : "[]",
"lastLedgerSwitchTimestamp" : "2019-09-03T04:00:02.255+09:00",
"state" : "Open",
"numberOfEntriesSinceFirstNotAckedMessage" : 1,
"totalNonContiguousDeletedMessagesRange" : 0,
"properties" : { }
}
}
}
```

**To Reproduce**
We suspect that reset-cursor operations which were executed right before caused the problem.
```
11:08 "reset-cursor (API)" was called. (11 days ago)
11:11 "reset-cursor (API)" was called. (1 minute ago)
```

The log at the time of operations is as follows.

Log:
```
11:08:08.171 [pulsar-web-26-32] INFO o.a.p.b.a.impl.PersistentTopicsBase - [] [persistent://] Received reset cursor on subscription to time 1566439687935
11:08:08.199 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.p.PersistentSubscription - [persistent://][] Unable to find position for timestamp 1566439687935. Resetting cursor to first position 25488301:0 in ledger
11:08:08.200 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.pulsar.broker.service.Consumer - Disconnecting consumer: Consumer{subscription=PersistentSubscription{topic=persistent://, name=}, consumerId=4, consumerName=7740a, address=/:50796}
11:08:08.200 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.AbstractDispatcherSingleActiveConsumer - Removing consumer Consumer{subscription=PersistentSubscription{topic=persistent://, name=}, consumerId=4, consumerName=7740a, address=/:50796}
11:08:08.200 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.p.PersistentSubscription - [persistent://][] Successfully disconnected consumers from subscription, proceeding with cursor reset
11:08:08.200 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.b.mledger.impl.ManagedCursorImpl - [] Initiate reset position to 25488301:0 on cursor
11:08:08.202 [BookKeeperClientWorker-OrderedExecutor-5-0] INFO o.a.b.mledger.impl.ManagedCursorImpl - [] reset position to 25488301:0 before current read position 26825736:4099 on cursor
11:08:08.202 [pulsar-web-26-32] INFO o.a.p.b.a.impl.PersistentTopicsBase - [serviceName1] [persistent://] Reset cursor on subscription to time 1566439687935
11:08:08.301 [ForkJoinPool.commonPool-worker-16] INFO o.a.pulsar.broker.service.ServerCnx - [/:50796] Subscribing on topic persistent:// /
11:08:08.301 [ForkJoinPool.commonPool-worker-16] INFO o.a.b.mledger.impl.ManagedCursorImpl - [-] Rewind from 25488301:0 to 25488301:0
11:08:08.301 [ForkJoinPool.commonPool-worker-16] INFO o.a.p.b.s.persistent.PersistentTopic - [persistent://][] Created new subscription for 4
11:08:08.301 [ForkJoinPool.commonPool-worker-16] INFO o.a.pulsar.broker.service.ServerCnx - [/:50796] Created subscription on topic persistent:// /

11:11:28.788 [pulsar-web-26-14] INFO o.a.p.b.a.impl.PersistentTopicsBase - [serviceName1] [persistent://] Received reset cursor on subscription to time 1567390228536
11:11:29.261 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.p.PersistentMessageFinder - [persistent://][] Found position 26825736:4100 closest to provided timestamp 1567390228536
11:11:29.261 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.pulsar.broker.service.Consumer - Disconnecting consumer: Consumer{subscription=PersistentSubscription{topic=persistent://, name=}, consumerId=4, consumerName=7740a, address=/:50796}
11:11:29.261 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.AbstractDispatcherSingleActiveConsumer - Removing consumer Consumer{subscription=PersistentSubscription{topic=persistent://, name=}, consumerId=4, consumerName=7740a, address=/:50796}
11:11:29.261 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.p.b.s.p.PersistentSubscription - [persistent://][] Successfully disconnected consumers from subscription, proceeding with cursor reset
11:11:29.261 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.b.mledger.impl.ManagedCursorImpl - [] Initiate reset position to 26825736:4100 on cursor
11:11:29.263 [BookKeeperClientWorker-OrderedExecutor-5-0] INFO o.a.b.mledger.impl.ManagedCursorImpl - [] reset position to 26825736:4100 skipping from current read position 25488301:7000 on cursor
11:11:29.263 [pulsar-web-26-14] INFO o.a.p.b.a.impl.PersistentTopicsBase - [serviceName1] [persistent://] Reset cursor on subscription to time 1567390228536
11:11:29.264 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.b.mledger.impl.ManagedLedgerImpl - [] End TrimConsumedLedgers. ledgers=46 totalSize=925996468
11:11:29.264 [bookkeeper-ml-workers-OrderedExecutor-22-0] INFO o.a.b.mledger.impl.ManagedLedgerImpl - [] Removing ledger 25488301 - size: 37740040
11:11:29.363 [ForkJoinPool.commonPool-worker-32] INFO o.a.pulsar.broker.service.ServerCnx - [/:50796] Subscribing on topic persistent:// /
11:11:29.363 [ForkJoinPool.commonPool-worker-32] INFO o.a.b.mledger.impl.ManagedCursorImpl - [-] Rewind from 26825736:4100 to 26825736:4100
11:11:29.363 [ForkJoinPool.commonPool-worker-32] INFO o.a.p.b.s.persistent.PersistentTopic - [persistent://][] Created new subscription for 4
11:11:29.363 [ForkJoinPool.commonPool-worker-32] INFO o.a.pulsar.broker.service.ServerCnx - [/:50796] Created subscription on topic persistent:// /
```

The partitioned topic has 16 partitions, but only one partition has a problem.

**Pulsar version**
broker: 2.3.2

Start with ManagedCursorImpl at the referenced messagesConsumedCounter logic, then inspect ManagedLedgerImpl.ENTRIES_ADDED_COUNTER_UPDATER and the partitioned-stats API path. Use the reset-cursor logs and cursor statistics to reproduce the mismatch; done means msgBacklog is zero when the cursor positions show no backlog, including for the affected partition.

Written by the indexing model from the issue text.