While working on ARO-13685 I (accidentally) crashed the CVO payload init containers. 

I found that the removal logic based on plain "rm" is not idempotent, so if any of the init containers crash mid-way, the restart will never be able to succeed.

The fix is to use "rm -f" in all places instead.

4.18 / main, but existed in prior versions    

always    

    1. inject a crash in the bootstrap init container https://github.com/openshift/hypershift/blob/99c34c1b6904448fb065cd65c7c12545f04fb7c9/control-plane-operator/controllers/hostedcontrolplane/cvo/reconcile.go#L353 

    2. the restarting previous init container "prepare-payload" will crash loop on "rm" not succeeding as the previous invocation already deleted all manifests
  
    

the prepare-payload init container will crash loop forever, preventing the container CVO from running

a crashing init container should be able to restart gracefully without getting stuck on file removal and eventually run the CVO container

based off the work in https://github.com/openshift/hypershift/pull/5315