Make fuzz_threaded_connections more robust

In `fuzz_threaded_connections`, if one thread is being run while
another is starved, and the running thread manages to call
`timer_tick_ocurred` twice after the starved thread constructs the
inbound connection but before it delivers the first bytes, we'll
receive an immediate error and `unwrap` it, causing failure.

The fix is trivial, simply remove the unwrap and return if we're
already disconnected when we do the initial read.

While we're here, we also reduce the frequency of the
`timer_tick_ocurred` calls to give us a chance to occasionally
deliver some additional messages.

Fixes #2073

Author: TheBlueMatt

lightning/src/ln/peer_handler.rs

@@ -2334,7 +2334,7 @@ mod tests {
 					let addr_b = NetAddress::IPv4{addr: [127, 0, 0, 1], port: 1001};
 					let initial_data = peers[1].new_outbound_connection(id_a, fd_b.clone(), Some(addr_a.clone())).unwrap();
 					peers[0].new_inbound_connection(fd_a.clone(), Some(addr_b.clone())).unwrap();
-					assert_eq!(peers[0].read_event(&mut fd_a, &initial_data).unwrap(), false);
+					if peers[0].read_event(&mut fd_a, &initial_data).is_err() { break; }
 
 					while start_time.elapsed() < std::time::Duration::from_secs(1) {
 						peers[0].process_events();
@@ -2364,8 +2364,10 @@ mod tests {
 								},
 							});
 
-						peers[0].timer_tick_occurred();
-						peers[1].timer_tick_occurred();
+						if ctr % 2 == 0 {
+							peers[0].timer_tick_occurred();
+							peers[1].timer_tick_occurred();
+						}
 					}
 
 					peers[0].socket_disconnected(&fd_a);