uv: implement a way to get client's ip/port.

[luqmana]

Can now get the remote ip address and port of a client upon accepting.

Quick example:

extern mod std;

use core::comm;
use std::uv;
use std::net::{ip, tcp};
use std::uv::*;

fn main() {

    let iotask = uv::global_loop::get();
    let addr = ip::v4::parse_addr("0.0.0.0");
    let port = 4000;

    tcp::listen(addr, port, 128u, iotask,
        |kill_ch| {
            info!("Listening on %s:%u.", ip::format_addr(&addr), port);
        },
        |conn, kill_ch| {
            do task::spawn_sched(task::ThreadPerCore) {
                match tcp::accept(conn) {
                    result::Ok(move sock) => {

                        let addr = sock.getpeername();
                        info!("accept %s:%u", ip::format_addr(&addr),
                                              ip::get_port(&addr));

                    }
                    result::Err(move e) => {
                        comm::send(kill_ch, Some(e));
                    }
                }
            }
        }
    );

}

[luqmana]

I couldn't figure out a way to determine if ipv4 or ipv6 was in use hence the separate getpeername and getpeername6.
Update: see subsequent commit.

[brson]

get_peer_name (or get_peer_ip) might be more consistent with rust naming conventions for the function in net::tcp

[brson]

A regression test would be nice, though I realize that the existing tests are very dense and setting up a simple test to exercise getpeername does not appear simple.

[brson]

Here's a test: brson@eb154f3

@luqmana are you ok with changing the name to get_peer_ip?

[luqmana]

Yep, i'm ok with it. I'll update it in just a bit with the rename + the test.

[luqmana]

Actually, I think get_peer_addr would make more sense since getting the port from the "ip" seems awkward, i.e:

let ip = sock.get_peer_ip();
let port = ip::get_port(&ip);

vs

let addr = sock.get_peer_addr();
let port = ip::get_port(&addr);

[brson]

I rebased this onto incoming. Thanks!