Cache SSH output between discovery and reasoning

[rysweet]

Scout polls each VM twice: Phase 1 FleetTUI.refresh_all + Phase 3 gather_context. Cache the first poll and pass it to the reasoner.

Files to modify:

• src/amplihack/fleet/_cli_session_ops.py
• src/amplihack/fleet/fleet_tui.py
• src/amplihack/fleet/_session_gather.py

Requirements:

1. After Phase 1 polling, store raw SSH output per VM in a dict
2. Pass cached output to gather_context in Phase 3
3. gather_context should use cached output when available, only re-poll if cache is None
4. Cache should be scoped to a single scout run (not persisted)
5. All 918+ fleet tests must pass
6. Commit changes on this branch