graphic: 0.920
boot: 0.843
architecture: 0.758
device: 0.740
arm: 0.623
performance: 0.617
PID: 0.458
files: 0.410
ppc: 0.385
permissions: 0.380
semantic: 0.377
debug: 0.321
register: 0.275
socket: 0.264
TCG: 0.233
mistranslation: 0.168
vnc: 0.165
peripherals: 0.152
VMM: 0.151
kernel: 0.081
user-level: 0.070
i386: 0.041
network: 0.036
risc-v: 0.025
hypervisor: 0.021
KVM: 0.016
virtual: 0.012
assembly: 0.011
x86: 0.007

system_powerdown only works once
Description of problem:
When the guest is configured to sleep on power button events, something in the ACPI states are not restored coming out of resume.  The first call to `system_powerdown` succeeds, but the second after waking the system is rejected in `acpi_pm1_evt_power_down()` since `ar->pm1.evt.en` is zero coming out of the resume path.

There is probably something deeper (or perhaps in seabios?) since removing the test in that handler doesn't cause a second sleep either.
Steps to reproduce:
![image](/uploads/60876bde4027c42699f2edf936bd874d/image.png)
1. Boot a guest configured to sleep when it receives a power button event
2. `system_powerdown` from the monitor to tell it to sleep
3. `info status` to verify that it is suspended
4. Wake the guest, either with `system_wakeup` or moving the mouse or something
5. `system_powerdown` has no effect
Additional information:
This is using qemu-7.2.0 built from source with a Windows 10 guest and IGD GPU+audio passthrough.