fix: stop parliament accept starvation on outbound dial (0.9.22)

Peer dial used per-step read timeouts so a stuck handshake could block the
single-threaded daemon for tens of seconds; inbound hellos piled up unread
(CLOSE-WAIT). Use a wall-clock dial budget, drain accepts before each dial,
skip self, and only one side initiates idle polls.

Co-authored-by: Cursor <cursoragent@cursor.com>

tests/test_peer_auth.c

@@ -9,6 +9,7 @@
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <time.h>
 
 typedef struct {
   os_socket_t listen;
@@ -109,8 +110,22 @@ int main(void) {
   pthread_join(th, NULL);
   assert(!srv2.ok);
 
+  /* Wall-clock dial budget: listener that never accepts must fail fast. */
+  os_socket_t ls3 = os_net_listen("127.0.0.1", 0, err, sizeof(err));
+  assert(ls3 != OS_SOCKET_BAD);
+  int port3 = os_net_port(ls3);
+  struct timespec t0, t1;
+  clock_gettime(CLOCK_MONOTONIC, &t0);
+  assert(cst_peer_dial("127.0.0.1", port3, &ca, ska, realm.pk, 800, &sess, err,
+                       sizeof(err)) != 0);
+  clock_gettime(CLOCK_MONOTONIC, &t1);
+  long ms = (long)((t1.tv_sec - t0.tv_sec) * 1000 +
+                   (t1.tv_nsec - t0.tv_nsec) / 1000000);
+  assert(ms >= 400 && ms < 2500);
+
   os_net_close(ls);
   os_net_close(ls2);
+  os_net_close(ls3);
   os_net_cleanup();
   puts("test_peer_auth: OK");
   return 0;