Incremental scout with staleness cache

[rysweet]

Store last scout results in ~/.amplihack/fleet/last_scout.json. On next scout, only re-poll sessions that changed or haven't been checked in N minutes.

Files to modify:

• src/amplihack/fleet/_cli_session_ops.py
• src/amplihack/fleet/_scout_cache.py (new file)

Requirements:

1. Create _scout_cache.py with load_cache/save_cache/is_stale functions
2. Cache format: {vm_name: {last_checked: timestamp, status: str, data: dict}}
3. Default staleness threshold: 5 minutes (configurable)
4. On scout, skip VMs whose cached status hasn't expired
5. Always re-poll VMs in error/unknown state
6. Add --full flag to force full re-poll
7. All 918+ fleet tests must pass
8. Commit changes on this branch