We may leave kvm_arch_vcpu_put with the ipipe notifier still registered
in case shared MSRs are still active. If a VM preemption will hit us
after that point but before the user return notifier fired, we may run
kvm_arch_vcpu_put again from __ipipe_handle_vm_preemption. Do not rely
on this being harmless and rather use a flag to decide if the run is
needed.

Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>