Revert "hypervisor/kvm: reduce VMM overhead to 100MiB for containers"

This reverts commit 6f76179.

We notice increased number of customer issues on the latest EVE
versions, when QEMU process was killed by the OOM killer. So revert
the overhead to 600M value.

In the EVE mater the estimation algorithm is more sophisticated,
please take a look here:

   d88091b ("New formula to estimate VMM memory overhead (x86)")

The rule of thumb is to leave the overhead tuning to a customer using
the corresponding field on the controller side.

Signed-off-by: Roman Penyaev <[email protected]>

Author: rouming

pkg/pillar/hypervisor/kvm.go

@@ -464,11 +464,7 @@ func (ctx kvmContext) Task(status *types.DomainStatus) types.Task {
 
 // TBD: Have a better way to calculate this number.
 // For now it is based on some trial-and-error experiments.
-// Container limit is reduced to 100MiB.
 func minVMMOverhead(config types.DomainConfig) int64 {
-	if config.IsOCIContainer() {
-		return 100 << 20 // Mb in bytes
-	}
 	return 600 << 20 // Mb in bytes
 }