qemu/kvm are instantly SIGKILLed by systemd on shutdown, without wait.
Description of problem:
systemd assumes it cannot terminate qemu, and SIGKILLs it. Instantly.
Steps to reproduce:
1. Start qemu on a systemd managed host
2. Shutdown/Reboot
Additional information:
Nothing on qemu's own log, besides that it is starting a vnc server.

```plaintext
# journalctl -b -1
...
Sep 22 18:38:04 local kernel: kvm_amd: TSC scaling supported
Sep 22 18:38:04 local kernel: kvm_amd: Nested Virtualization enabled
Sep 22 18:38:04 local kernel: kvm_amd: Nested Paging enabled
Sep 22 18:38:04 local kernel: kvm_amd: Virtual VMLOAD VMSAVE supported
Sep 22 18:38:04 local kernel: kvm_amd: Virtual GIF supported
Sep 22 18:38:04 local kernel: kvm_amd: LBR virtualization supported
...
Sep 22 18:38:50 local systemd-logind[721]: The system will reboot now!
Sep 22 18:38:50 local systemd-logind[721]: System is rebooting.
Sep 22 18:38:50 local sddm-helper[850]: Signal received: SIGTERM
...
Sep 22 18:38:50 local systemd[1]: Stopping User Manager for UID 1000...
Sep 22 18:38:50 local systemd-logind[721]: Removed session 1.
Sep 22 18:38:50 local systemd[854]: Activating special unit Exit the Session...
Sep 22 18:38:50 local systemd[854]: app-org.kde.konsole-1ab3dac6a1db4b29b55899b477b32975.scope: Failed to kill control group /user.slice/user-1000.slice/user@1000.service/app.slice/>
Sep 22 18:38:50 local systemd[854]: app-org.kde.konsole-1ab3dac6a1db4b29b55899b477b32975.scope: Killing process 1708 (qemu-system-x86) with signal SIGKILL.
Sep 22 18:38:50 local systemd[854]: app-org.kde.konsole-1ab3dac6a1db4b29b55899b477b32975.scope: Killing process 1712 (kvm-nx-lpage-recovery-1708) with signal SIGKILL.
Sep 22 18:38:50 local systemd[854]: app-org.kde.konsole-1ab3dac6a1db4b29b55899b477b32975.scope: Failed to kill control group /user.slice/user-1000.slice/user@1000.service/app.slice/>
Sep 22 18:38:50 local systemd[854]: Stopped Konsole - Terminal.
... (some other applications terminanting normally )
Sep 22 18:38:50 local systemd[854]: app-org.kde.konsole-1ab3dac6a1db4b29b55899b477b32975.scope: Consumed 10.068s CPU time.
Sep 22 18:38:50 local systemd[854]: Removed slice User Background Tasks Slice.
Sep 22 18:38:50 local systemd[854]: background.slice: Consumed 2.960s CPU time.
...
```

I cannot explain why it sends SIGKILL to qemu/kvm... it is the same second as the shutdown started, their docs says there's a delay for that.

Also, other processes owned by the user received a single SIGTERM after qemu was SIGKILLed. Some even take a couple seconds to exit and are not SIGKILLed.