Implement shutdown, as required by mirage-flow 4.0.0

[hannesm]

Would be great to have a review on this.

close

From my understanding, in the UNIX sockets API the SO_LINGER is relevant (to specify how long close() may block) - since we don't have SO_LINGER, let's just care about the normal case (text paraphrased from https://www.cl.cam.ac.uk/~pes20/Netsem/alldoc.pdf):

• close sends any remaining data and gracefully closes the connection
• if close is called on a pre-established state: socket is closed and removed
• if on a listening socket: close and remove that socket, and aborts each of the connections on the socket's pending and completed connection queues.

If I understand, our flow is never a listening socket. Our unlisten eventually should care about the last point.

shutdown

Depending on `read | `write | `read_write, the specific half is shut down:

• read: shutting down the read-half empties the socket’s receive queue, but data will still be delivered to it and subsequent recv() calls will return data.
• write: Shutting down the write-half of a TCP connection causes the remaining data in the socket’s send queue to be sent and then TCP’s connection termination to occur.

TL;DR

There's some discrepancy from my intuition (and the man page shutdown(2) SHUT_RD and the model: according to the man page) "Further receives will be disallowed". In this PR, I push None to the Lwt_dllist -- with the hope that this leads to the expected behaviour (those who have knowledge about the internals should comment whether there's something else that needs to be done).

[djs55]

LGTM. I'm not sure exactly how the SHUTDOWN_RECEIVE is supposed to behave but I've never found a use for it. The SHUTDOWN_SEND to perform a half-close is really useful though and this looks good.

[hannesm]

Thanks @djs55. I'll go ahead with merge and release once CI is green.