Improve bounded mpsc documentation

[Darksonn]

This clarifies various things in the documentation for bounded mpsc channels.

[Darksonn]

This is tested by

tokio/tokio/tests/sync_mpsc.rs

Lines 275 to 301 in 672be92

	#[tokio::test]
	async fn recv_close_gets_none_reserved() {
	let (tx1, mut rx) = mpsc::channel::<i32>(1);
	let tx2 = tx1.clone();
	let permit1 = assert_ok!(tx1.reserve().await);
	let mut permit2 = task::spawn(tx2.reserve());
	assert_pending!(permit2.poll());
	rx.close();
	assert!(permit2.is_woken());
	assert_ready_err!(permit2.poll());
	{
	let mut recv = task::spawn(rx.recv());
	assert_pending!(recv.poll());
	permit1.send(123);
	assert!(recv.is_woken());
	let v = assert_ready!(recv.poll());
	assert_eq!(v, Some(123));
	}
	assert!(rx.recv().await.is_none());
	}

[Darksonn]

This particular note is what inspired this PR, since whether the channels works when one end is sync and the other is async is a surprisingly common question.

[LucioFranco]

sleep feels weird here but I can't think of anything better.