Send keep-alive pings to prevent server idle timeout killing sessions

[PureWeen]

Problem

The headless server has an unconfigurable ~35 minute idle timeout. When a session is resumed but not actively used, the server kills it with session.shutdown.

Observed in: ci-agentic session - resumed at 19:38:39, server killed at 20:13:25 (34m 46s idle after resume).

Root Cause

The headless server tracks client inactivity per-session. When no messages flow from the client for ~35 minutes, the server shuts down the session. There is no --session-idle-timeout flag.

Proposed Fix

Send lightweight keep-alive pings every ~10 minutes for sessions PolyPilot considers in use.

Considerations

• Need to identify the right SDK method for a keep-alive
• Should NOT interfere with active tool execution or trigger model inference
• May be an upstream server bug

Related

• PR Fix multi-agent worker failures, session persistence, server health, and history recovery #391 added watchdog session.shutdown detection (commit ccee665)
• Issue Upstream: Copilot CLI headless server breaks when global CLI cleans shared native module directory #392 (posix_spawn upstream bug)
• Issue Spinner vanishes briefly during premature idle recovery cycles #395 (spinner gap during premature idle recovery)