Added timeout when disconnecting connector in Windows

[nikhalim]

No description provided.

[ericgross2]

This might have some side effects. I think there is a supported pattern where you can call send() and then disconnect and be guaranteed that the sends are completed before the disconnect happens. By adding this short timeout it might cause data to be lost in this case.

I’m a little unclear what causes the original problem. Is it possible there is queued send data?

[nikhalim]

That's what I'm afraid of when adding the timeout. However, I couldn't think of another way to get the EventHandler thread to exit when Disconnect is called. For some odd reason, this situation occurs only when RDMA Close is called on Windows's listener session and Linux's delays closing its connector session. I could not reproduce this situation with both Windows, both Linux nor Windows connecting to Linux.

Is there a way for me to identify whether there's any queued send data?