The kubevirt_vmi_launcher_memory_overhead_bytes does not report the virt-launcher overhead bytes accurately.
It is also not reported for all vmis.

Example YAML of a VMI that this metric is not reported for - https://drive.google.com/drive/folders/1nFQP8d1L3mWwuJgXEWTEA2nWUcuJYK16?usp=sharing


Seems like it is also reporting values not only for the running VMIs.

1. Create a single VM in a namespace
2. 
3.

When we calculate 

(sum(kube_pod_resource_request{resource="memory", namespace="example-namespace"})-sum(kubevirt_vmi_launcher_memory_overhead_bytes{namespace="example-namespace"}) )/1024/1024/1024

For a VM that I asked 1GB of memory I expected to get 1GB, but I got 1.0335276126861572 GB.
Which means the overhead is not accurate.

The metrics should report the overhead accuratly.