fix(tcp): MUX模式网口失联 — 对端关闭时用tcp_abort替代tcp_close避免TIME_WAIT耗尽pcb池

根因: tcp_close()将对端关闭的pcb推入TIME_WAIT(120s), 占用MEMP_TCP_PCB池(仅4个),
多连接同时断开后pcb池耗尽, tcp_new()返回NULL, 新连接无法建立直到120s超时释放。

核心修复:
- tcp_server/client: 对端关闭(p=NULL)时tcp_abort替代tcp_close, pcb立即释放
- ch390_runtime: PKT_ERR恢复强制OR上RCR_RXEN(与WCH官方一致)
- ch390_runtime: TX连续超时3次自动emergency reset
- ch390_runtime: 每5秒health_check读VID验证芯片存活
- main: App_StartLinksIfNeeded失败时不标记g_links_started, 允许重试
- main: MUX逐帧RTT printf改为#if DEBUG门控, 减少主循环延迟
- uart_trans: MUX帧解析改为先搜0x7E再消费header, 非法帧只丢1字节

App/tcp_client.c

@@ -52,13 +52,11 @@ static err_t tcp_client_on_recv(void *arg, struct tcp_pcb *pcb, struct pbuf *p,
         tcp_recv(pcb, NULL);
         tcp_sent(pcb, NULL);
         tcp_err(pcb, NULL);
-        if (tcp_close(pcb) != ERR_OK) {
-            tcp_abort(pcb);
-        }
+        tcp_abort(pcb);
         ctx->pcb = NULL;
         ctx->status.state = TCP_CLIENT_STATE_DISCONNECTED;
         ctx->next_retry_ms = HAL_GetTick() + ctx->config.reconnect_interval_ms;
-        return ERR_OK;
+        return ERR_ABRT;
     }
 
     for (q = p; q != NULL; q = q->next) {