fix: remove retry_timer, use waiter-based round-robin on send_q for fairness

- tcp_proxy_server: drain_cb uses etcp_router_on_send_ready instead of retry_timer
- socks_proxy: on_read_cb + tx_waiter_cb use etcp_router_on_send_ready instead of waiter_register
- Result: FIFO round-robin via ll_queue waiters on send_q, all streams get fair share
- No timers — wakeup is event-driven via queue_data_get → check_waiters
- TCP backpressure: read_queue callback_suspended prevents over-reading

tests/test_socks_http_proxy.c

@@ -304,13 +304,13 @@ static int worker_main(int wtype, const char* socks_proxy, const char* http_prox
         unlink(f1); unlink(out);
 
         snprintf(cmd, sizeof(cmd),
-                 "curl -s -I --connect-timeout 10 --max-time 30 -x http://%s %s/f_1k.bin 2>/dev/null | head -1 | grep -q '200 OK'",
+                 "curl -s -I --connect-timeout 15 --max-time 30 -x http://%s %s/f_1k.bin 2>/dev/null | head -1 | grep -q '200 OK'",
                  http_proxy, url_http);
         if (system(cmd) != 0) { fprintf(stderr, "[FAIL] %s HEAD 200\n", tname); return 1; }
 
         snprintf(out, sizeof(out), "%s/opt_%d.txt", tmpdir, getpid());
         snprintf(cmd, sizeof(cmd),
-                 "curl -s -i --connect-timeout 10 --max-time 30 -x http://%s -X OPTIONS -o %s %s/f_1k.bin 2>/dev/null",
+                 "curl -s -i --connect-timeout 15 --max-time 30 -x http://%s -X OPTIONS -o %s %s/f_1k.bin 2>/dev/null",
                  http_proxy, out, url_http);
         if (system(cmd) != 0) { fprintf(stderr, "[FAIL] %s OPTIONS curl\n", tname); return 1; }
         snprintf(cmd, sizeof(cmd), "grep -q 'Allow:' %s", out);