feat: use two gauges for suspend and reboot

[Keyruu]

This PR deprecates the single comin_host_info gauge (which used labels for host state) and adds two separate gauges comin_is_suspended and comin_need_to_reboot.

Why?

Labels aren't ideal for frequently changing values. Prometheus labels are designed for filtering and aggregating, not for values that change often (for example whether a node is suspended). Using them this way causes two problems:

• Stale time series - Prometheus creates a new time series for every unique label combination and only cleans up old ones after a certain interval. This means outdated label combinations linger and can produce misleading results when querying.
• Incorrect semantics - A Gauge is meant to represent a value that goes up and down. By splitting into separate Gauges, each metric is a single, stable time series that simply updates its value.