Schedule tasks where we've seen but not processed PTRACE_EVENT_EXIT

If we see the PTRACE_EVENT_EXIT for a task while running a different
task in unlimited-ticks mode in `Scheduler::reschedule`, it looks like
nothing ever actually calls `handle_ptrace_exit_event` on it, and so
nothing ever PTRACE_CONT's the task out of the exit-stop and into the
zombie state.

This seems to manifest itself as rr not reaping processes properly when
they receive asynchronous core-dumping signals (e.g. SIGSEGV sent by
`raise` or `kill`).

Fix this issue by checking if there's a pending PTRACE_EVENT_EXIT to
deal with on the task in `Scheduler::is_task_runnable`, and allowing the
task to be executed if so.

Fixes #3882

src/Scheduler.cc

@@ -349,7 +349,10 @@ bool Scheduler::is_task_runnable(RecordTask* t, WaitAggregator& wait_aggregator,
     }
   }
 
-  if (t->waiting_for_ptrace_exit && !t->was_reaped()) {
+  if (t->seen_ptrace_exit_event() && !t->handled_ptrace_exit_event()) {
+    LOGM(debug) << "  " << t->tid << " has a pending PTRACE_EVENT_EXIT to process; we can run it";
+    return true;
+  } else if (t->waiting_for_ptrace_exit && !t->was_reaped()) {
     LOGM(debug) << "  " << t->tid << " is waiting to exit; checking status ...";
   } else if (t->is_stopped() || t->was_reaped()) {
     LOGM(debug) << "  " << t->tid << "  was already stopped with status " << t->status();