The metairq sample sometimes failed due to stuck because it has a
potential race when recording the latencies. The previous solution
checks whether the stats.mirq greater than MAX_EVENT. But stats.mirq
might also have a race here. So we change to check the msg sequence,
its correctness ensured by the kernel's message queue.
And also remove the spinlock in previous fix because tests only use
1 CPU, no need to add this. It also improves the fix for #40889,
to prevent test failed sporadically on acrn_ehl_crb and ehl_crb.
Fix#42874.
Signed-off-by: Enjia Mai <enjia.mai@intel.com>
8300ffd302