CHROMIUM: Revert "pm: re-program timer HW only when the exit latency is not zero"

This reverts commit 3a58885.

That change sometimes causes the EC's timing to behave incorrectly after
some time in a low-power mode, such that the EC's view of time runs
about a million times too fast. On some devices this reliably causes
watchdog resets, while on others it's more subtle (often manifesting as
a loss of battery communication).

Nuvoton is investigating possible fixes, but until better understanding
of the problem is obtained this revert should prevent it from occurring.

BUG=b:372418669,b:372724562,b:361773754,b:369897523
TEST=Manual testing on sundance is unable to reproduce the issue,
     where it previously occurred about 50% of the time after leaving
     the system in S0ix overnight.

Change-Id: Ib0f9b8097ec42a126f3f3f4e4505273e83ab6b72
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/third_party/zephyr/+/5942743
Commit-Queue: YH Lin <[email protected]>
Reviewed-by: caveh jalali <[email protected]>
Tested-by: YH Lin <[email protected]>

Author: tari

subsys/pm/pm.c

@@ -170,8 +170,7 @@ bool pm_system_suspend(int32_t ticks)
 	}
 #endif
 
-	if ((z_cpus_pm_state[id].exit_latency_us != 0) &&
-	    (ticks != K_TICKS_FOREVER)) {
+	if (ticks != K_TICKS_FOREVER) {
 		/*
 		 * We need to set the timer to interrupt a little bit early to
 		 * accommodate the time required by the CPU to fully wake up.