Add non-blocking semantic indexing with Jobs Dashboard UI

[illAssad]

Summary

This PR fixes a critical issue where semantic indexing blocks the entire server, making it unresponsive until indexing completes. The fix implements background processing with real-time progress tracking.

Problem

When running semantic indexing on large codebases (2000+ files), the server became completely unresponsive because the async def api_index_semantic endpoint called blocking functions (ingest_file() with Ollama embeddings) synchronously, blocking FastAPI's event loop.

Solution

• Background Processing: Use FastAPI's BackgroundTasks to run semantic indexing in a separate thread
• Job Tracking: Add INDEXING_JOBS dictionary with thread-safe progress updates
• New Endpoints:
  • GET /api/jobs - List all indexing jobs
  • GET /api/jobs/{job_id} - Get specific job status
• Jobs Dashboard UI: New React component with live progress bars, status indicators, and auto-refresh
• Backward Compatibility: background=false parameter for sync mode when needed

Changes

• mcp_server/server.py - Background worker, job tracking, new endpoints
• mcp_server/claude_code_mcp.py - Updated tool schema with background parameter
• frontend/src/components/JobsDashboard.tsx - New dashboard component
• frontend/src/pages/MainApp.tsx - Added Jobs tab
• tests/test_mcp_server.py - Updated tests

Testing

• Server returns job_id in ~15ms (previously blocked for minutes)
• Server stays responsive during indexing
• Progress updates visible in real-time via Jobs tab
• Completed jobs retained for 24 hours, auto-cleanup prevents memory leaks

Screenshots

The Jobs Dashboard shows:

• Summary cards (Running, Queued, Completed, Failed)
• Per-job progress bars with animations
• Status badges with appropriate colors
• Timestamps and duration tracking
• Error messages for failed jobs