[Feature Request] Add ability to resume idle agents from /agents picker with Running/Recent tabs

Bug Description
Feature request: resume idle agents from the /agents picker

Today /agents only lists live tool loops — once an agent returns its final report and goes idle, it disappears from the picker. The agent's
conversation context is preserved on the runtime side and is reachable via SendMessage from an orchestrator agent, but end-users have no way to
resume it from the CLI directly.

Why this matters: in long sessions with many parallel sub-agents (4+ background dispatches per turn), the orchestrator becomes a bottleneck.
The user wants to "open" a finished agent in a side conversation to query its thinking, ask follow-ups, or have it run additional tests —
without polluting the orchestrator's context window or having the orchestrator transcribe every back-and-forth.

Proposed UX: /agents shows two tabs (or one combined list with status badges) — Running and Recent. Selecting an idle agent re-opens its
transcript and lets the user type directly to it (same as if it were running). The agent resumes its tool loop on first message and exits again
on its next final report.

Bonus: an /agent <id> slash command for keyboard-driven jump-to-agent.

This was raised after a session with 4 background Agent-tool agents where the user wanted to chat with each individually and lost the
affordance once they completed.

Environment Info

• Platform: win32
• Terminal: null
• Version: 2.1.136
• Feedback ID: 2e9fa80d-7443-4ec8-ad97-bca8e4a89694