[feature request] custom runtimeClassName

[Archimonde666]

Often with kubernetes cluster in hybrid environments where GPU and CPU nodes exists, it is not easily feasible to make nvidia runtimeClassName as the default one.

As of today, i digged through it and it doesn't appear to be a mechanism implemented nor in the schemas or in the interactive services charts to account for the runtimeClassName override.

I guess it would be best done by implementing two new schema file allowing for different runtimeclass specified for CPU and GPU deployments as it is done with ressources and ressources-gpu