fix: use UDP connect to probe local address

[seven-mile]

Background

ENET, which is based on UDP, requires a local address to bind and receive datagrams. Previously, we populated and recorded the local address during the RTSP handshake, which is usually bound to a TCP socket. While this approach generally works well, I discovered that in certain network environments—such as when using "aTrust VPN"—the discrepancy between TCP and UDP (ENET) behavior causes issues (see issue #1300).

The root cause is that some VPN software redirects TCP traffic to a loopback proxy for forwarding. This leads to unexpected behavior when calling socket APIs:

>>> host, port = "10.xx.xx.xxx", 48010
>>> sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
>>> sock.connect((host, port))
>>> local_addr = sock.getsockname()[0]
# Returns 127.0.0.1
>>> local_hostname = socket.gethostbyaddr(local_addr)[0]
# Returns localhost.sangfor.com.cn

While this behavior does not break the TCP connection itself, it causes failures when we apply the local address retrieved from a TCP socket to a UDP socket.

Approach

The solution is to use a UDP socket (SOCK_DGRAM) to probe the address instead. udp_sock.getsockname returns the correct local interface address in these environments.

Since the exact implementation of the RTSP handshake can vary, I prefer performing the local address resolution immediately after STAGE_NAME_RESOLUTION, once the remote address is available.

Note

As this repository serves as a common library for various Moonlight projects, I do not have the necessary environment to cover every possible code path, platform, and network configuration. Any guidance on conducting more thorough or automated tests would be greatly appreciated.