rfkill: don't use [delayed_]work_pending()

There's no need to test whether a (delayed) work item in pending
before queueing, flushing or cancelling it.  Most uses are unnecessary
and quite a few of them are buggy.

Remove unnecessary pending tests from rfkill.  Only compile
tested.

Signed-off-by: Tejun Heo <tj@kernel.org>
Acked-by: Johannes Berg <johannes@sipsolutions.net>
Cc: "John W. Linville" <linville@tuxdriver.com>
Cc: linux-wireless@vger.kernel.org

net/rfkill/input.c

@@ -148,11 +148,9 @@ static unsigned long rfkill_ratelimit(const unsigned long last)
 
 static void rfkill_schedule_ratelimited(void)
 {
-	if (delayed_work_pending(&rfkill_op_work))
-		return;
-	schedule_delayed_work(&rfkill_op_work,
-			      rfkill_ratelimit(rfkill_last_scheduled));
-	rfkill_last_scheduled = jiffies;
+	if (schedule_delayed_work(&rfkill_op_work,
+				  rfkill_ratelimit(rfkill_last_scheduled)))
+		rfkill_last_scheduled = jiffies;
 }
 
 static void rfkill_schedule_global_op(enum rfkill_sched_op op)