Keep managed-agent status polling off the UI thread

[neodaysoff]

What changed

Run list_managed_agent_runtimes through Tauri's blocking worker pool instead of executing its filesystem, process-inspection, locking, and persistence work inline in the command handler.

Why

The desktop polls this command frequently. With 15 managed agents on an ARM64 NVIDIA DGX Spark running Ubuntu/GNOME X11, the synchronous command caused the UI to stop servicing input and Mutter _NET_WM_PING requests. GNOME then presented the "Buzz is not responding" / Force Quit dialog.

Moving the existing implementation to spawn_blocking keeps the desktop IPC/UI path responsive without changing the returned status data or lifecycle behavior.

Impact

• Prevents managed-agent status polling from freezing the desktop UI.
• Preserves the existing command contract for frontend callers.
• Applies to all desktop platforms; no DGX-specific behavior is introduced.

Validation

• cargo fmt --manifest-path desktop/src-tauri/Cargo.toml -- --check
• cargo check --manifest-path desktop/src-tauri/Cargo.toml on Linux ARM64 against current main
• Rebuilt and exercised Buzz Desktop on a DGX Spark with 15 managed-agent listeners; the app remained responsive through repeated status polls and normal navigation where the previous build consistently triggered GNOME's Force Quit warning.