[Bug] Autorecovery pod crashes repeatedly while scaling up zookeeper

### Search before asking

- [X] I searched in the [issues](https://github.com/apache/pulsar/issues) and found nothing similar.

### Version

Running Pulsar 3.0.0 in Kubernetes

### Minimal reproduce step

Install a basic pulsar cluster then scale up zookeeper.

### What did you expect to see?

Maybe some reconnects in bookkeeper and autorecovery.

### What did you see instead?

The autorecovery pod crashed and restarted while the zookeepers were scaling and bookkeeper was reconnecting.
```
2023-07-28T16:16:19,281+0000 [main] ERROR org.apache.bookkeeper.client.BookieWatcherImpl - Failed to get bookie list :
org.apache.bookkeeper.client.BKException$ZKException: Error while using ZooKeeper
at org.apache.bookkeeper.discover.ZKRegistrationClient.lambda$getChildren$4(ZKRegistrationClient.java:352) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.zookeeper.ZooKeeperClient$25$1.processResult(ZooKeeperClient.java:1177) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:668) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:553) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
Caused by: org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /ledgers/available
at org.apache.zookeeper.KeeperException.create(KeeperException.java:118) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.zookeeper.KeeperException.create(KeeperException.java:54) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.bookkeeper.discover.ZKRegistrationClient.lambda$getChildren$4(ZKRegistrationClient.java:351) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
... 3 more
2023-07-28T16:16:19,294+0000 [main] ERROR org.apache.bookkeeper.replication.AutoRecoveryMain - Failed to build AutoRecovery Server
java.io.IOException: Failed to create bookkeeper client
at org.apache.bookkeeper.replication.Auditor.createBookKeeperClient(Auditor.java:105) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.replication.AutoRecoveryMain.(AutoRecoveryMain.java:94) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.server.service.AutoRecoveryService.(AutoRecoveryService.java:40) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.replication.AutoRecoveryMain.buildAutoRecoveryServer(AutoRecoveryMain.java:371) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.replication.AutoRecoveryMain.doMain(AutoRecoveryMain.java:339) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.replication.AutoRecoveryMain.main(AutoRecoveryMain.java:316) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
Caused by: org.apache.bookkeeper.client.BKException$ZKException: Error while using ZooKeeper
at org.apache.bookkeeper.discover.ZKRegistrationClient.lambda$getChildren$4(ZKRegistrationClient.java:352) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.zookeeper.ZooKeeperClient$25$1.processResult(ZooKeeperClient.java:1177) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:668) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:553) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
Caused by: org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode = NoNode for /ledgers/available
at org.apache.zookeeper.KeeperException.create(KeeperException.java:118) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.zookeeper.KeeperException.create(KeeperException.java:54) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.bookkeeper.discover.ZKRegistrationClient.lambda$getChildren$4(ZKRegistrationClient.java:351) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.bookkeeper.zookeeper.ZooKeeperClient$25$1.processResult(ZooKeeperClient.java:1177) ~[org.apache.bookkeeper-bookkeeper-server-4.16.1.jar:4.16.1]
at org.apache.zookeeper.ClientCnxn$EventThread.processEvent(ClientCnxn.java:668) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
at org.apache.zookeeper.ClientCnxn$EventThread.run(ClientCnxn.java:553) ~[org.apache.zookeeper-zookeeper-3.8.1.jar:3.8.1]
```

### Anything else?

_No response_

### Are you willing to submit a PR?

- [ ] I'm willing to submit a PR!

Start with the stack-trace entry points in AutoRecoveryMain.java, Auditor.java, and ZKRegistrationClient.java, then reproduce scaling ZooKeeper while BookKeeper reconnects in Kubernetes. Investigate the NoNodeException for /ledgers/available; done means the autorecovery pod remains healthy instead of repeatedly crashing during this transition.

Written by the indexing model from the issue text.