fix(namespaces): increase QPS and burst rate limits

[nobbs]

This PR increases the rate limits (QPS and burst) of the namespace switching component. This fixes the issue of the namespace switcher timing out while fetching the namespaces from a cluster due to cache discovery of the kubernetes API / client-go taking too long on clusters with lots of CRDs. This is a known issue, cf. this blog post from 2022 and other users have also reported issues for similar timeouts in #127.

I've updated QPS to 50.0 and burst to 300 from the defaults specified in

kubeswitch/vendor/k8s.io/client-go/rest/config.go

Lines 43 to 46 in a6981cf

	const (
	DefaultQPS float32 = 5.0
	DefaultBurst int = 10
	)

The default burst for cache discovery is set in

kubeswitch/vendor/k8s.io/client-go/discovery/discovery_client.go

Line 688 in a6981cf

func setDiscoveryDefaults(config *restclient.Config) error {

to 300, but I'm not sure if it really is being properly propagated here.

Anyway, the new values 50/ 300 are based on the defaults in the kubectl code found in https://github.com/kubernetes/kubernetes/blob/5f594f42159f2ee84f805d382638f77dd5967177/staging/src/k8s.io/kubectl/pkg/cmd/cmd.go#L95-L97

Setting this substantially increased the responsiveness and I haven't since seen this issue on clusters with ~350 CRDs where the namespacer switcher previously was not able to grab namespaces at all.

[danielfoehrKn]

Thank you very much for the PR. Could you try to rebase it on master - golangci-lint is failing for this build. Master contains an updated version of golangci-lint. Thank you!

[nobbs]

Thank you very much for the PR. Could you try to rebase it on master - golangci-lint is failing for this build. Master contains an updated version of golangci-lint. Thank you!

Right, sorry - apparently I had a previous fork and forgot to pull in changes. Done, should pass the ci workflow now.