propolis-cli -s is very confused with a port number

in propolis-server we accept a SocketAddr as a binary argument outright, meaning you can run ./propolis-server run path/to/ovmf.fd 0.0.0.0:12345 and you're off.

however, in propolis-cli we have -s for the server address, and -p for the port. this makes for some confusing UI: if you ./propolis-cli -s 127.0.0.1:12345 to talk to that server you just started, this actually just gets you

failed to lookup address information: node name or service name not known

... because, i think, resolve_host is taking the string from the server parameter and forcing Rust to treat it as literally either IP address or hostname, with an explicit port 0 for the socket. :3000 makes it clearly not an IP address so Rust's ToSocketAddrs falls back to trying to look up the host which it also won't do, and whoops name or service not known.

i think we should probably drop the -p flag and expect ip:port in all cases.

Start in bin/propolis-cli/src/main.rs, especially the argument handling and resolve_host at lines 466-473; compare it with the SocketAddr handling in bin/propolis-server/src/main.rs at lines 81-82. The change is complete when propolis-cli accepts a server address in ip:port form without the separate port confusion and no longer produces the lookup error for addresses such as 127.0.0.1:12345.