Port `fleet` observe-agent monitoring from the Go fork

[beengud]

Port fleet observe-agent monitoring from the Go fork

Source: beengud/observe PR #3. Files: cmd_fleet.go.
New CLI status: Absent.

Old command surface

observe fleet status   [--window <duration>]   # current agent inventory: host, platform env, version, auth result, instance ID (newest first)
observe fleet host <hostname> [--window <duration>]  # event history for one host incl. agent start time (restarts)
observe fleet versions [--window <duration>]   # version distribution across the fleet
observe fleet auth     [--window <duration>]   # auth-check status from all agents; failures sorted first, incl. HTTP code + auth URL

--window accepts Go duration strings (20m, 1h, 24h, 168h); anchored at now, extends backward. 20m recommended for live inventory (heartbeat ~every 10 min).

Behavior / API

• Queries the Default.Observe Agent/Events resource dataset (heartbeat events) via OPAL query. In the new CLI this can build on the existing query execution path.

Acceptance

• All four subcommands ported with --window and --format json/csv; output columns match the old tool.