sched/pthread: fix race condition on pthread_cond_wait()

pthread_cond_wait() should be an atomic operation in the mutex lock/unlock.
Since the sched_lock() has been wrongly deleted in the previous commit,
the context switch will occurred after the mutex was unlocked:

--------------------------------------------------------------------
  Task1(Priority 100)             |      Task2(Priority 101)
                                  |
  pthread_mutex_lock(mutex);      |
  |                               |
  pthread_cond_wait(cond, mutex)  |
  |  |                            |
  |  |                            |
  |  ->enter_critical_section()   |
  |  ->pthread_mutex_give(mutex)  | ----> pthread_mutex_lock(mutex);    // contex switch to high priority task
  |                               |       pthread_cond_signal(cond);    // signal before wait
  |                               | <---- pthread_mutex_unlock(mutex);  // switch back to original task
  |  ->pthread_sem_take(cond->sem)|                                     // try to wait the signal, Deadlock.
  |  ->leave_critical_section()   |
  |
  |  ->pthread_mutex_take(mutex)  |
  |                               |
  pthread_mutex_lock(mutex);      |
---------------------------------------------------------------------

This PR will bring back sched_lock()/sched_unlock() to avoid context switch to ensure atomicity

Signed-off-by: chao an <anchao@xiaomi.com>

sched/pthread/pthread_condwait.c

@@ -94,6 +94,7 @@ int pthread_cond_wait(FAR pthread_cond_t *cond, FAR pthread_mutex_t *mutex)
       sinfo("Give up mutex / take cond\n");
 
       flags = enter_critical_section();
+      sched_lock();
       mutex->pid = INVALID_PROCESS_ID;
 #ifndef CONFIG_PTHREAD_MUTEX_UNSAFE
       mflags     = mutex->flags;
@@ -116,6 +117,7 @@ int pthread_cond_wait(FAR pthread_cond_t *cond, FAR pthread_mutex_t *mutex)
           ret = status;
         }
 
+      sched_unlock();
       leave_critical_section(flags);
 
       /* Reacquire the mutex.