\GatewayClient\Gateway::sendToGroup在高并发场景中很难确保发送成功

最近在一个大流量活动中使用了Workerman/GatewayWorker，发现在后台群推消息时，偶发会出现部分接收方无法接收消息的问题。

日志中表现为：


大概就是在

```
} // 运行在其它环境中，通过注册中心得到gateway地址
else {
$addresses = static::getAllGatewayAddressesFromRegister();
foreach ($addresses as $address) {
$gateway_data['ext_data'] = isset($address_connection_array[$address]) ?
json_encode(array('group'=> $group, 'exclude'=> $address_connection_array[$address])) :
$default_ext_data_buffer;
static::sendToGateway($address, $gateway_data);
}
}
```

这里获取得到gateway地址后循环调用发送。但是如果循环过程中有一个gateway挂了，后面就无法发送了。

Start at Gateway::sendToGroup and inspect the shown branch that calls getAllGatewayAddressesFromRegister() and static::sendToGateway($address, $gateway_data). Trace how a failed gateway affects the loop and how failures are reported. Done means a failed gateway does not prevent messages from being attempted for the remaining gateway addresses, with the behavior verified under the reported high-concurrency scenario.

Written by the indexing model from the issue text.