Fix shutdown bugs by holding loggers in shared_ptrs

[jupp0r]

This change fixes a shutdown crash occuring when using an
external io_service to drive a websocketpp server. The crash
itself is reproduced by the following code:

int main(int argc, const char * argv[])
{
    boost::asio::io_service ioService;
        std::unique_ptr<boost::asio::io_service::work> ioServiceWork(new boost::asio::io_service::work(ioService));
            std::thread ioServiceThread(boost::bind(&boost::asio::io_service::run, &ioService));

    {
        auto server = std::unique_ptr<websocketpp::server<websocketpp::config::asio>>(new websocketpp::server<websocketpp::config::asio>());
        server->init_asio(&ioService);
        server->listen(0);
        server->start_accept();

        server->stop();
    }

    ioServiceWork.reset();
    ioService.stop();
    ioServiceThread.join();

    return 0;
}

The cause of the crash is a dangling pointer to alog or elog objects in endpoints.
This is mitigated by storing logging objects as shared_ptrs. This ensures that
they are kept alive and able to log even if their owning structures have already
been deallocated.

[jupp0r]

@zaphoyd: any updates?