Webhook wrongly patches "request" into a shared pool user's manifest

[Levovar]

Describe the bug
Mutating webhook shall only add limits to a Pod's CPU resource field when a Pod asks for shared pool devices, but shall not add requests.

Reasoning
Limits are only added to the Pod to make default Kubernetes logic provision CFS quotas for the requested shared CPU slices.
However, by also adding request to the Pod manifests' comes with a major side-effect. K8s does not know that the requesting Pod will not actually use the Node's Kubelet allocatable CPU resources, and will unnecessarily decrease it with the patched amount.
This effectively decreases the usable capacity of the default pool when a shared pool using Pod is scheduled to the Node.

To Reproduce
Steps to reproduce the behavior:

1. Create shared and default pool on a Node
2. Decrease Kubelet Node Allocatable CPU capacity to match the size of the default pool
3. Create a Pod asking for e.g. 200ms slice from the Node's shared pool

Expected behavior
3. Upon the instantiation of the Pod Kubelet recognized Noda Allocatable CPU pool remains the same, and is not decreased with 200ms

Additional info
Long term we should probably set the quotas on our own to avoid presenting an unrealistic picture on the Kubernetes interfaces. In short term setting limit via K8s can be still okay as it has a much smaller effect on the state of the cluster compare to setting request.