rtpengine: Fix possible startup crash in notification process

When the "notification_sock" modparam was set, the process responsible
for notifications would start by scanning a non-NULL terminated string
past the end boundary, sometimes leading to a crash on startup.

Fixes #2947

(cherry picked from commit cb59a84)

modules/rtpengine/rtpengine.c

@@ -3506,7 +3506,7 @@ static void rtpengine_notify_process(int rank)
 	union sockaddr_union ss;
 	char buffer[RTPENGINE_DGRAM_BUF];
 
-	p = strrchr(rtpengine_notify_sock.s, ':');
+	p = q_memchr(rtpengine_notify_sock.s, ':', rtpengine_notify_sock.len);
 	if (!p) {
 		LM_ERR("no port specified in notification socket %.*s!\n",
 				rtpengine_notify_sock.len, rtpengine_notify_sock.s);

modules/rtpengine/rtpengine.h

@@ -31,7 +31,7 @@
 
 struct rtpe_node {
 	unsigned int		idx;			/* overall index */
-	str					rn_url;			/* unparsed, deletable */
+	str					rn_url;			/* unparsed, deletable, NULL-term */
 	int					rn_umode;
 	char				*rn_address;	/* substring of rn_url */
 	int					rn_disabled;	/* found unaccessible? */