feat(cli): support custom SSH port for remote gateway deployment

[bokerh]

Problem Statement

openshell gateway add ssh://user@host:8080 parses 8080 as the gateway's HTTPS port. There is no way to specify a custom SSH port (e.g. 2222) for the connection itself. The CLI strips the SSH connection down to user@host, which forces the underlying Docker library to default to port 22.

Proposed Design

1. CLI Flag: Add --ssh-port <PORT> to remote commands (gateway add, gateway start, etc).
2. Internal API: Add ssh_port: Option<u16> to RemoteOptions.
3. Execution: In create_ssh_docker_client, append the port to the bollard connection string if set (ssh://user@host:PORT).

Alternatives Considered

• ~/.ssh/config alias: Users define a Host with a custom Port. Rejected: This is currently broken. resolve_ssh_hostname in metadata.rs only parses the hostname line from ssh -G and ignores the port line, so the custom port is lost before the connection starts.
• URI overloading: Support ssh://user@host:2222:8080. Rejected: Invalid URL format.

Agent Investigation

• crates/openshell-cli/src/run.rs L925: URL parser assigns the parsed port to the gateway.
• crates/openshell-bootstrap/src/lib.rs L59: RemoteOptions struct has no ssh_port field.
• crates/openshell-bootstrap/src/docker.rs L240: create_ssh_docker_client passes ssh://user@host to bollard without a port.
• crates/openshell-bootstrap/src/metadata.rs L164: resolve_ssh_hostname ignores the port field from ssh -G output.

Checklist

• I've reviewed existing issues and the architecture docs
• This is a design proposal, not a "please build this" request

[drew]

It's a bit confusing that the ssh:// URI takes a gateway port in the first place. Let me look to see if we can infer the gateway port from the ssh host. That way the ssh URI can more conventionally specify the SSH port from the URI.

For example openshell gateway add ssh://user@host:2222 would connect over ssh on port 2222 and automatically grab the gateway running on 8080.