QNetworkAccessManager hangs indefinitely when loading a corrupted cache file

When making a cached request for which there is a corrupted cache file (in particular a cache file with empty contents), the request never terminates.

Steps to reproduce

The issue can be reproduced with the following MWE:

#include <QtCore>
#include <QtNetwork>int main(int argc, char *argv[])
{
    QCoreApplication a(argc, argv);    QNetworkAccessManager nam;
    QObject::connect(&nam, &QNetworkAccessManager::finished, &nam, [](QNetworkReply *reply) {
        QTextStream(stdout) << "Got a reply: " << reply->url().toDisplayString() << Qt::endl;
        reply->deleteLater();
    });
    auto cache = new QNetworkDiskCache;
    cache->setCacheDirectory("/tmp/qtcache");
    nam.setCache(cache);
    nam.setRedirectPolicy(QNetworkRequest::ManualRedirectPolicy);    QNetworkRequest req(QUrl("https://www.qt.io/hubfs/Quality_Assurance.png"));
    auto reply = nam.get(req);
    QObject::connect(reply, &QNetworkReply::finished, &a, &QCoreApplication::quit);    return a.exec();
} 

1. Run the code once to download a cached copy of https://www.qt.io/hubfs/Quality_Assurance.png (this is just an URL that has proper caching headers set) in `/tmp/qtcache`. It should print `Got a reply: https://www.qt.io/hubfs/Quality_Assurance.png` then exit cleanly.
2. Manually corrupt the cache file that got created in `/tmp/qtcache/data8/` to only keep the first line (the exact name of the cache file may be different on your machine):

   $ sed -i -n '1p' /tmp/qtcache/data8/6/3ia6ms7v.d

1. Run the code again. It now never prints the `Got a reply` message and hangs forever.

Additional Information

I tracked this down to `QNetworkReplyHttpImplPrivate::_q_cacheLoadReadyRead` failing early because `cacheLoadDevice->bytesAvailable()` is `0` in that case, but `_q_cacheLoadReadyRead` is never called again afterwards.

Note that QNetworkDiskCache seems to be using atomic rename in `QNetworkDiskCachePrivate::storeItem` so from my understanding it should not be able to create such corrupted/partial cache files yet we saw it happen here.