sched/signal: fix the occasional issue where SIGCONT was not executed

In multi-core scenarios, a problem occurs: the target thread receives SIGCONT but does not execute them.

Further analysis reveals that in the sig_dispatch function, the target thread is placed in the readytorun queue only if the SIGCONT signal is sent and the target thread is in the TSTATE_TASK_STOPPED state. However, if the following conditions occur:

CPU0				CPU1
dispatch SIGSTOP
dispatch SIGCONT(check state)
				sched_suspend(set task_state = TSTATE_TASK_STOPPED)

This causes the target thread on CPU 1 to receive SIGCONT but not be placed in the readytorun queue, remaining in the stopped state.
Therefore, A flag is needed to indicate whether the SIGCONT action has been executed. At the same time, the critical section protection range of nxsig_stop_task needs to be expanded to prevent another thread from executing the SIGCONT logic during the execution of this function.

Signed-off-by: wangzhi16 <wangzhi16@xiaomi.com>

sched/sched/sched_suspend.c

@@ -36,6 +36,7 @@
 
 #include "sched/sched.h"
 #include "sched/queue.h"
+#include "signal/signal.h"
 
 #ifdef CONFIG_SMP
 /****************************************************************************
@@ -102,11 +103,24 @@ void nxsched_suspend(FAR struct tcb_s *tcb)
 {
   irqstate_t flags;
   bool switch_needed;
+  FAR sq_entry_t *entry;
 
   DEBUGASSERT(tcb != NULL);
 
   flags = enter_critical_section();
 
+  /* Check if received SIGCONT */
+
+  sq_for_every(&tcb->sigpendactionq, entry)
+    {
+      FAR sigq_t *sigq = (FAR sigq_t *)entry;
+      if (sigq->info.si_signo == SIGCONT)
+        {
+          leave_critical_section(flags);
+          return;
+        }
+    }
+
   /* Check the current state of the task */
 
   if (tcb->task_state >= FIRST_BLOCKED_STATE &&