allow remote-triggered stream resets after closing a send stream

[marten-seemann]

I think I understand why we made this decision back then. Take a look at the send stream states from RFC 9000:

It explicitly contains the transition Data Sent to Reset Sent. While it makes sense to have CancelWrite after Close a no-op (it's basically a misuse of the API, you might argue), a sender might receive a STOP_SENDING from the peer. In that case, it does make sense to send a RESET_STREAM and stop retransmitting the stream data.

Originally posted by @marten-seemann in #4404 (comment)