[enhancement] Implement session persistence and recovery for interrupted executions

[github-actions]

Background

The Copilot SDK includes session persistence capabilities that allow SDK applications to list and reconnect to existing Copilot sessions. Recent SDK updates have enhanced this with:

1. Session context tracking (v0.1.24) - SessionContext with working directory, git root, repository, and branch info
2. Session filtering (v0.1.24) - listSessions() can filter by context fields in PR #427 by Jeremy Moseley
3. Improved session management - Better object disposal and lifecycle handling in PR #481 by Steve Sanderson

These capabilities enable applications to reconnect to orphaned sessions after crashes or interruptions.

Related SDK commits:

• Expose session context, add filtering, and context_changed event
• Fix ObjectDisposedException when disposing session after client.StopAsync()

Proposal

Implement session persistence and recovery in Planeteer to allow resuming interrupted executions:

1. Session ID tracking: Store Copilot session IDs in the persisted plan when execution starts
2. Graceful shutdown: On normal exit or Ctrl+C, properly dispose sessions and mark tasks as interrupted
3. Recovery on restart: When loading a plan with incomplete tasks, detect orphaned sessions using listSessions() with repository context filtering
4. Resume option: Offer users the choice to resume from orphaned sessions or start fresh execution
5. Session cleanup: Provide a way to dispose orphaned sessions that are no longer needed

Implementation points:

• Extend Task model with optional sessionId?: string field
• Save plan JSON after each task completion (incremental persistence)
• On load, check for tasks with status: 'running' and associated session IDs
• Use listSessions({ filter: { repository, branch } }) to find orphaned sessions
• Add "Resume" option in execute screen if orphaned sessions are detected

Benefit

Session persistence makes Planeteer more robust and production-ready:

• Crash recovery: Users can resume work after unexpected crashes, network issues, or terminal closures
• Long-running executions: For plans with many tasks, users can stop and resume execution across multiple sessions
• Resource management: Prevents accumulation of orphaned Copilot sessions consuming resources
• Professional UX: Matches user expectations from CI/CD and build tools that support resumable operations
• Debugging: Users can inspect the state of failed executions and retry specific tasks

Acceptance Criteria

• Add sessionId?: string field to Task interface in src/models/plan.ts
• Update src/services/executor.ts to store session IDs when creating agent sessions
• Implement incremental plan persistence - save plan JSON after each task completes or fails
• On plan load, detect tasks with status: 'running' and check if their sessions still exist using listSessions()
• Display orphaned sessions in execute screen with options: "Resume", "Start Fresh", "Clean Up"
• Implement graceful shutdown handler (SIGINT/Ctrl+C) to dispose active sessions and mark tasks as interrupted
• Add session cleanup utility to dispose orphaned sessions by repository context
• Update SDK dependency to ^0.1.24 or later to ensure session context filtering is available
• Add unit tests for session recovery logic
• Document session persistence behavior in README

AI generated by Weekly Enhancement Suggestions