[FEAT] Suspend tests should examine time in hardware sleep state

Where is 90% come from? any specification can support it?
And what happen if the test less than 60 seconds, still support 90% of the time?

Empirical measurements. I've found this this is a good amount of time for hardware to settle into proper steady state.

In shorter cycles you spend more of the time entering and exiting and so your residency is low.

For example I find 10s cycle only reaches 75% of the time.

There might be Windows HLK requirements for this too, but if so they don't appear public.

I have consulted with our certificate team regarding the test criteria for Windows HLK requirements for this case, but unfortunately, no one was able to provide an answer. As a result, we will put this to wishlist and wait until we obtain more information about the test requirements.

(Follow up from https://bugs.launchpad.net/fwts/+bug/2029452/comments/7)

> I don't know much about /sys/power/suspend_stats/last_hw_sleep, could you provide more details?
> I am concerning the slp_s0_residency_usec will only be increased when the platform actually go to PC10.
> That's intel suggests to test the s2idle. https://github.com/intel/S0ixSelftestTool
> How about /sys/power/suspend_stats/last_hw_sleep, will it be increased on other PC states or only PC10?

So this file is "opt-in". It will only be visible if it is populated by the OEM's driver.
For AMD this is amd_pmc driver: https://github.com/torvalds/linux/blob/v6.5-rc7/drivers/platform/x86/amd/pmc.c#L376
For Intel this is intel_pmc driver: https://github.com/torvalds/linux/blob/v6.5-rc7/drivers/platform/x86/intel/pmc/core.c#L1317

For Intel case yes s0ix counter only increases while you're in PC10.