[#380] Track chatbot semantic search costs and performance

[sahilds1]

Add Semantic Search Usage Tracking

Overview

Implements telemetry for semantic search operations by introducing a new SemanticSearchUsage model that captures performance metrics and search statistics without impacting the main search workflow.

Benefits

• Performance insights: Identify slow encoding or database query operations
• Usage analytics: Track search patterns by user and over time
• Result quality: Monitor distance distributions to assess embedding model effectiveness
• Zero-impact: Failures in telemetry collection don't affect user-facing search results

Key Implementation Details

1. Unauthenticated User Support

• User field includes explicit documentation:
  help_text="User who performed the search (null for unauthenticated users)"
• Service logic: user=user if (user and user.is_authenticated) else None
• Enables tracking of anonymous search patterns while maintaining data integrity

2. Cascade Deletion

• User relationship configured with on_delete=models.CASCADE
• Rationale: Search usage records are considered disposable when the associated user account is deleted
• Prevents orphaned telemetry data accumulation

3. Resilient Telemetry

• Database writes wrapped in try/except block to ensure telemetry failures never break search functionality
• Error logging for debugging: "Failed to create semantic search usage database record: {e}"
• Records created even for zero-result searches (with num_results_returned=0 and None distance stats)
• Warning logged when searches return no results for operational visibility