修复SpkDownloadMgr重大bug

1. 原有依靠Qt Network的连接超时不可靠，改为自己使用Watchdog值跟踪。
2. 原有代码在请求时大量未对Reply设置正确的workerId,已修正
3. 原有代码不能自动在出错worker出错但其他worker已经全部完成时进行得分配，已修正
4. 原有代码ActiveWorkerCount计算方法不准确，每次link都会增加，但是由于重试重分配都会Link而且Worker不会从
List中删除，所以已改为任务开始时一次加满
5. 原有代码忘记在WorkerFinish时检查有否未写入的cache以及未关闭下载文件
6. 原有代码未将Reply的errorOccurred信号与WorkerError槽连接，导致这个槽并没有发挥什么卵用

测试文件改为有道词典，小一点
增加了等待HEAD请求时光标变成忙的特性

inc/spkdownload.h

@@ -7,6 +7,9 @@
 /**
  * @note SpkDownloadMgr does NOT do download scheduling and other things; it's only a multithreaded
  *       downloader; it manages the threads that are downloading stuff from the Internet.
+ *
+ *       Because of this, SpkDownloadMgr does not support complex download queues, cannot handle
+ *       pauses, and can only work on a sequential list of tasks.
  */
 
 class SpkDownloadMgr : public QObject
@@ -34,15 +37,22 @@ class SpkDownloadMgr : public QObject
      *        unnecessary race conditions and data safety problems.
      *        DownloadWorker is also used in mFailureRetryQueue to indicate the blocks that needed
      *        to be retried on other servers.
+     *
+     *        Each worker has a watch dog value, incremented each time the download speed is
+     *        updated, and zeroed each time the worker has data ready. If the value exceeds a
+     *        preset maximum, then this worker is considered timed out and killed.
      */
     struct DownloadWorker
     {
       QNetworkReply *Reply; ///< Reply from the network
+      int Watchdog; ///< Watch dog value watching for a timed out worker
       qint64 BeginOffset; ///< Where should a worker start downloading
       qint64 BytesNeeded; ///< How many bytes a worker should fetch in total
       qint64 BytesRecvd; ///< How many bytes a worker has received till now
     };
 
+    constexpr static int WatchDogMaximum = 7;
+
     struct RemoteFileInfo
     {
       qint64 Size = -1;