Add last_ledger_closed_ago_seconds to metrics and root endpoint

[bartekn]

core_latest_ledger - history_latest_ledger doesn't always determine if Horizon is fully sync with the network. It's possible that underlying Core instance is behind or stuck. A new metric with a number of seconds since last known ledger close time would better explain this.