State polling logic should be unified

[cachebag]

We have several different connection-progress loops, with duplicate error logic:

• Wait for activation
• Handle config timeout
• Handle failure reasons
• Detect stuck disconnected
• Detect stuck config

We can unify into something like

async fn wait_for_connection_state(
    &self,
    dev: &NMDeviceProxy<'_>,
    timeout_ms: u64,
) -> Result<()>;