Currently, to deploy custom commits of the GPU Operator, we rely on the helm charts of the repository. This method was satisfactory to test the behavior of the operator code, however the resources deployed by helm and OLM will be different.

To properly test the metrics being generated by the operator and shared with OpenShift, we need to be able to deploy the OLM bundle image of the operator.

In addition, this will allow a better testing on OpenShift of the GPU Operator master branch before its release, as it will use the exact operand images that will be published in the certified catalog.