Handle additional `UiStartOrder` messages better

[dnwiebe]

Currently, when the Daemon receives a UiStartOrder message, it checks to see if the Node is running. If it isn't, the Daemon starts it and returns a UiStartResponse message containing the Node's process ID and port. But if the Node is already running, the Daemon returns an error that contains nothing but an error code and a message.

This error is not helpful for a UI that wants to connect to an already-running Node.

Ultimate Intended Situation: We want the Daemon to stop responding with an error when the Node is already running. Here's what the UiStartResponse message looks like now--the message that the Daemon sends when the Node was not running and has just been started up:

pub struct UiStartResponse {
    #[serde(rename = "newProcessId")]
    pub new_process_id: u32,
    #[serde(rename = "redirectUiPort")]
    pub redirect_ui_port: u16,
}

We'd like the Daemon to send a message that looks like the following, whether the Node has just been started up or whether it was already started and running:

pub struct UiStartResponseV2 {
    #[serde(rename = "wasRunning")]
    pub was_running: bool,
    #[serde(rename = "processId")]
    pub process_id: u32,
    #[serde(rename = "uiPort")]
    pub ui_port: u16,
}

Actual errors, of course, should still result in error responses; but sending a message like this when the Node is already running will provide the information necessary to connect to it.

Big Problem:
At the moment, all the user interfaces are expecting a UiStartResponse, not a UiStartResponseV2. If we stop sending the former and instead send the latter, they'll all break.

Audacious Solution:
Therefore: initially, let's send both UiStartResponse and UiStartResponseV2 in response to a UiStartOrder message. Whichever the UI isn't expecting should be ignored, perhaps with a warning message, and everything ought to work fine. Once all the UIs are expecting UiStartResponseV2, we can stop sending the UiStartResponse, and stop getting warning messages.