[FLINK-36876] Add proxy for restClient

[chenyuzhi459]

What is the purpose of the change

Add proxy for restClient, which supports to use shared EventLoopGroup

Brief change log

• Introduce a new class RestClientProxy as the proxy of restClient
• Introduce a new configuration: kubernetes.operator.flink.client.io.threads to specify the total thread of sharded eventLoopGroup
• RestClientProxy supports to use sharded eventLoopGroup

Verifying this change

Unit tests + manual verification

Does this pull request potentially affect one of the following parts:

• Dependencies (does it add or upgrade a dependency): (no)
• The public API, i.e., is any changes to the CustomResourceDescriptors: (no)
• Core observer or reconciler logic that is regularly executed: (yes)

Documentation

• Does this pull request introduce a new feature? (no)
• If yes, how is the feature documented? (docs)

[gyfora]

Makes sense on a high level, reflection is not great but it's okay if we have a better solution in newer Flink versions.

Which Flink version will contain the Fix for this?

[chenyuzhi459]

It depends on when will the pr apache/flink#25788 could be merged.

If necessary, I can submit the pr fix in both 1.20 and 2.x versions of Flink.

@gyfora

[chenyuzhi459]

Is there any other problem with this pr? I don't know why CI check was cancelled

@gyfora

[gyfora]

Added a minor clarifying question, also please rebase to retrigger the CI

[gyfora]

How will this actually solve the original problem? The rest client still creates an individual group (that we close here) won't that cause the same issue anyways?

[chenyuzhi459]

Because PoolThreadCache is a request-thread-related cache, it is created and initialized when a request is made using the NioEventLoopGroup;

here it is created and closed immediately, with no requests in between, so there is no PoolThreadCache leakage issue.

[chenyuzhi459]

Could you help take a look at the latest CI failure issue, which doesn't seem to be related to this code commit?