* Use "mizu" namespace instead of "default". Create and delete as necessary.

* Wait until namespace is deleted.

* Distinguish between timeout and other errors.

* Sorted consts.

* k8s provider gets the names of Mizu serviceaccount, clusterrole and clusterrolebindings from caller.

* Renames.

* Remove non-namespaced mizu resources when finished: clusterrole and clusterrolebindings.

* Don't wait for namespace deletion if it was already deleted.

* When watching pods, check for cancellation before reading from channels.

* Allow user to cancel resource deletion and to cancel the wait.

* Increased cleanup timeout.

* go mod tidy.

* Ignore cli build products.

* Print err.

* Don't delete clusterrole and clusterrolebinding if we do not have permissions.

* Added roles list in README.

* Added clusterrole and clusterrolebindings examples.