sched/bmq: Fix nice level mis-behavior

This is caused by previous fix code for torvalds#89, new forked tasks doesn't
deboost quickly enough and impact run time of existed tasks with same
priority.

boost/deboost strategy need further adjustment, but rule is: task should
be deboost when renew time slice.

kernel/sched/bmq.h

@@ -77,8 +77,7 @@ static inline void sched_update_rq_clock(struct rq *rq) {}
 
 static inline void sched_task_renew(struct task_struct *p, const struct rq *rq)
 {
-	if (rq_switch_time(rq) > sysctl_sched_base_slice)
-		deboost_task(p);
+	deboost_task(p);
 }
 
 static inline void sched_task_sanity_check(struct task_struct *p, struct rq *rq) {}