More predictable Downloader::startDownload()

Before this change `Downloader::startDownload()` might avoid starting a
new download when a download with the specified URI was already present
in its cache. This might be confusing for the following reasons:

1. uri is not the only parameter of `Downloader::startDownload()` - a
   target download directory may also be specified through the second
   `options` parameter. Thus calling `Downloader::startDownload()` twice
   with the same URI but different download directories would not save
   files into the second directory.

2. Files of a completed download may be removed, whereupon downloading the same
   files again won't be a no-op. However in such a situation
   `Downloader` refuses to actually repeat a previous download.

src/downloader.cpp

@@ -169,12 +169,6 @@ std::vector<std::string> Downloader::getDownloadIds() const {
 std::shared_ptr<Download> Downloader::startDownload(const std::string& uri, const std::vector<std::pair<std::string, std::string>>& options)
 {
   std::unique_lock<std::mutex> lock(m_lock);
-  for (auto& p: m_knownDownloads) {
-    auto& d = p.second;
-    auto& uris = d->getUris();
-    if (std::find(uris.begin(), uris.end(), uri) != uris.end())
-      return d;
-  }
   std::vector<std::string> uris = {uri};
   auto gid = mp_aria->addUri(uris, options);
   m_knownDownloads[gid] = std::make_shared<Download>(mp_aria, gid);