[LibOS] Add SIGTERM signal to list of pending signals

Previously, `get_all_pending_signals()` helper function did not check
for the SIGTERM signal (which has a special handling via
`g_host_injected_signal`). Syscalls like `rt_sigsuspend()` and
`rt_sigtimedwait()` use this helper function. Because of that erroneous
behavior, apps like Nginx would wait for signals on one of these
syscalls but would *not* react to SIGTERM (i.e., won't react to `kill
<app>`). This commit fixes this bug, which in turn fixes nginx and
ra-tls-nginx examples.

Signed-off-by: Dmitrii Kuvaiskii <dmitrii.kuvaiskii@intel.com>

CI-Examples/ra-tls-nginx/run.sh

@@ -10,7 +10,8 @@ sleep 10
 test "$(curl --insecure -s https://localhost:8000/)" = "$(cat html/index.html)"
 ret=$?
 
-kill -9 $pid
+kill $pid
+wait $pid
 
 if test $ret -eq 0
 then

libos/src/bookkeep/libos_signal.c

@@ -151,6 +151,11 @@ void get_all_pending_signals(__sigset_t* set) {
 
     __sigemptyset(set);
 
+    int host_sig = __atomic_load_n(&g_host_injected_signal, __ATOMIC_RELAXED);
+    if (host_sig && host_sig != 0xff) {
+        __sigaddset(set, host_sig);
+    }
+
     if (__atomic_load_n(&current->pending_signals, __ATOMIC_ACQUIRE) == 0
             && __atomic_load_n(&g_process_pending_signals_cnt, __ATOMIC_ACQUIRE) == 0) {
         return;