[documentation request] add example: watch customresource (and handle timeouts)

[lknite]

Describe the bug There seems to be some inconsistency around watching a CRD, including an example would help to resolve confusion.

Inconsistency

• Watch will fail after a timeout, what's the right way to handle this to get things going again?
• Will the example below fully relist all items before watching, or will it pick back up where it left off using resourceVersion? (documenting an example with a comment letting us know will inform all future folks who want to watch a CRD instance)
• If the below doesn't pick back up, how to implement such that it does?
• I found an example which uses resourceVersion (though not with a custom resource), maybe we need to do something similiar in csharp? Or does using generic like below handle picking back up with the last resourceVersion automatically? (https://www.baeldung.com/java-kubernetes-watch)

Example

            var generic = new GenericClient(kubeclient, group, version, plural);

            while (true)
            {
                try
                {
                    await foreach (var (type, item) in generic.WatchNamespacedAsync<CrdInvite>("gge"))
                    {
                        Console.WriteLine("(event) [" + type + "] " + plural + "." + group + "/" + version + ": " + item.Metadata.Name);
                    }
                }
                catch (Exception ex)
                {
                    Console.WriteLine("Exception occured while performing 'watch': " + ex);
                }
            }

In the kubernetes documentation it suggests handling the watch timeout like so, how does the kubernetes client library do it?

[IvanJosipovic]

Check out this repo, it's a informer implementation. If you put a loop around the Start code it should do what you want.

https://github.com/microsoft/reverse-proxy/tree/main/src/Kubernetes.Controller/Client

[lknite]

I think there may a "most effective" way to write this code, but not sure what that'd be. In anycase, everything's working good enough for now to close out this ticket.