Consider removing unnecessary reference copies in prologues to closures in AutorecoveringConnection

[antiduh]

Before forming a closure, a couple bits of code in AutorecoveringConnection save off a second copy of the this pointer:

       public void BeginAutomaticRecovery()
        {
            lock (recoveryLockTarget)
            {
                if (!performingRecovery)
                {
                    performingRecovery = true;
>>>             var self = this;

                    recoveryTaskFactory.StartNew(() =>
                    {
                        if (!self.ManuallyClosed)
                        {
                            try
                            {
#if NETFX_CORE
                                System.Threading.Tasks.Task.Delay(m_factory.NetworkRecoveryInterval).Wait();
#else
                                Thread.Sleep(m_factory.NetworkRecoveryInterval);
#endif
                                self.PerformAutomaticRecovery();

And:

        private void Init(IFrameHandler fh)
        {
            m_delegate = new Connection(m_factory, false,
                fh, this.ClientProvidedName);

>>>     AutorecoveringConnection self = this;
            EventHandler<ShutdownEventArgs> recoveryListener = (_, args) =>
            {
                lock (recoveryLockTarget)
                {
                    if (ShouldTriggerConnectionRecovery(args))
                    {
                        try
                        {
                            self.BeginAutomaticRecovery();

I don't believe these copies serve any purpose.

[michaelklishin]

I am responsible for connection recovery in most clients and have a theory :) This was a more or less literate port of the Java client implementation, and a final this reference is required there. Feel free to submit a PR that does away with them 👍