sched/nohz: Remove unnecessarily complex error handling pattern from find_new_ilb()

find_new_ilb() returns nr_cpu_ids on failure - which is the usual
cpumask bitops return pattern, but is weird & unnecessary in this
context: not only is it a global variable, it it is a +1 out of
bounds CPU index and also has different signedness ...

Its only user, kick_ilb(), then checks the return against nr_cpu_ids
to decide to return. There's no other use.

So instead of this, use a standard -1 return on failure to find an
idle CPU, as the argument is signed already.

Signed-off-by: Ingo Molnar <[e-mail]>
Reviewed-by: Joel Fernandes (Google) <[e-mail]>
Link: https://lore.kernel.org/r/[e-mail]

kernel/sched/fair.c

···
 			return ilb_cpu;
 	}
 
-	return nr_cpu_ids;
+	return -1;
 }
 
 /*
···
 		nohz.next_balance = jiffies+1;
 
 	ilb_cpu = find_new_ilb();
-
-	if (ilb_cpu >= nr_cpu_ids)
+	if (ilb_cpu < 0)
 		return;
 
 	/*