Improve low-level IO monitor functions (for IOCP and kqueue)

[njsmith]

These should take a queue, rather than returning a queue, to support the case where you want to register multiple filters and multiplex them onto the same queue. (This is probably more important for kqueue than IOCP, b/c in IOCP you can take the completion key and use it for multiple registrations? but either way.)

CompletionKeyEventInfo should also carry the actual completion key, b/c it becomes ambiguous.

[oremanj]

We also need a replacement for the deprecated UnboundedQueue.

[richardsheridan]

Maybe I'm missing something, but haven't we been able to resolve this for some time? Currently the only internal usages of UnboundedQueue only need to meet the requirements of:

• Have a put_nowait method
• don't raise

Specifically:

trio/src/trio/_core/_io_windows.py

Line 613 in c951d95

queue.put_nowait(info)

and

trio/src/trio/_core/_io_kqueue.py

Line 97 in c951d95

receiver.put_nowait(event)

we could swap this for send_nowait and have the monitor functions accept a MemorySendChannel, although we'd be better off to protect the internal state of the IOManager with a try/except/warn?/pass in case the user forgets to make it unbounded.

We could also be even more flexible and have these methods just accept a callback. I know callbacks are "bad," but then we don't have to concern ourselves with how the user wants to handle things.