Enable TCP_NODELAY optimization

It looks like `TCP_NODELAY` was disabled due to the bug fixed in the
previous commit. Enabling it substantially improves keep-alive
performance with `ab`:

Before:

```
Time per request:       0.272 [ms] (mean)
```

After:

```
Time per request:       0.033 [ms] (mean)
```

TODO

@@ -1,3 +1 @@
-- keepalive performance against ab sucks
-- keepalive+TCP_NODELAY hangs!
 - send headers using sendfile syscall - fewer packets

darkhttpd.c

@@ -817,13 +817,11 @@ static void init_sockin(void) {
                    &sockopt, sizeof(sockopt)) == -1)
         err(1, "setsockopt(SO_REUSEADDR)");
 
-#if 0
     /* disable Nagle since we buffer everything ourselves */
     sockopt = 1;
     if (setsockopt(sockin, IPPROTO_TCP, TCP_NODELAY,
             &sockopt, sizeof(sockopt)) == -1)
         err(1, "setsockopt(TCP_NODELAY)");
-#endif
 
 #ifdef TORTURE
     /* torture: cripple the kernel-side send buffer so we can only squeeze out