[Bug] issues 15271 Refactoring the bug in the registry center

### Pre-check

- [x] I am sure that all the content I provide is in English.

### Search before asking

- [x] I had searched in the [issues](https://github.com/apache/dubbo/issues?q=is%3Aissue) and found no similar issues.

### Apache Dubbo Component

Java SDK (apache/dubbo)

### Dubbo Version

dubbo 3.3 branch. jdk 11

### Steps to reproduce this issue

Configure multiple ZooKeeper registries, with one service not started and configured as check=false

### What you expected to happen

There is an issue with this approach. When multiple ZooKeeper registries are configured and one of them has check=false, the logic in org.apache.dubbo.remoting.zookeeper.curator5.Curator5ZookeeperClient#Curator5ZookeeperClient was changed from the original behavior of closing the connection when !connected to closing it only when check && !connected. As a result, the subsequent workflow incorrectly assumes that the connection has been successfully established and continues writing node data to that connection.

A typical example is node creation in org.apache.dubbo.remoting.zookeeper.curator5.AbstractZookeeperClient#create. Each recursive attempt to create the node fails, which prolongs the application startup time. Eventually, an exception is thrown in org.apache.dubbo.remoting.zookeeper.curator5.Curator5ZookeeperClient#createPersistent(java.lang.String, boolean), causing the application startup to fail.

### Anything else

I want to solve this issue, I'm willing to submit a PR. Could you provide me with some ideas on how to approach this problem?
My approach is to add a registry != null check to all methods in org.apache.dubbo.registry.ListenerRegistryWrapper, while also passing in the original URL object. When org.apache.dubbo.registry.ListenerRegistryWrapper#getUrl is invoked, if registry == null, it should return the original URL directly. In this way, when the following code is executed in org.apache.dubbo.registry.integration.RegistryDirectory#subscribe:
registry.getUrl()
.getParameter(
RegistryConstants.REGISTRY_CLUSTER_KEY,
registry.getUrl().getParameter(PROTOCOL_KEY));
it will no longer trigger a NullPointerException.

### Do you have a (mini) reproduction demo?

- [ ] Yes, I have a minimal reproduction demo to help resolve this issue more effectively!

### Are you willing to submit a pull request to fix on your own?

- [x] Yes I am willing to submit a pull request on my own!

### Code of Conduct

- [x] I agree to follow this project's [Code of Conduct](https://www.apache.org/foundation/policies/conduct)

Start by reproducing the multiple-ZooKeeper-registry case on the dubbo 3.3 branch with one unavailable registry configured as check=false. Read Curator5ZookeeperClient#Curator5ZookeeperClient, AbstractZookeeperClient#create, Curator5ZookeeperClient#createPersistent, ListenerRegistryWrapper, and RegistryDirectory#subscribe. Done means startup no longer treats the unavailable registry as successfully connected or repeatedly fails while creating node data.

Written by the indexing model from the issue text.