[PECOBLR-1143] Implement telemetry Phase 4-5: Export infrastructure and opt-in configuration

[samikshya-db]

Summary

This PR implements the remaining core infrastructure for PECOBLR-1143 (Phases 4-5), completing the foundational telemetry system for the Databricks SQL Go driver.

Phase 4: Export Infrastructure ✅

• Implemented telemetryExporter with HTTP POST to /api/2.0/telemetry-ext
• Added retry logic with exponential backoff (100ms base, 3 retries)
• Integrated with circuit breaker for endpoint protection
• Implemented tag filtering via shouldExportToDatabricks()
• Added comprehensive error swallowing
• Support for both HTTP/HTTPS URLs (testing + production)

Phase 5: Opt-In Configuration Integration ✅

• Implemented isTelemetryEnabled() with 5-level priority logic:
  1. forceEnableTelemetry=true - bypasses all server checks (testing/internal)
  2. enableTelemetry=false - explicit opt-out (always disabled)
  3. enableTelemetry=true + server flag - user opt-in with server control
  4. Server flag only - default Databricks-controlled behavior
  5. Default disabled - fail-safe default
• Integrated with existing featureFlagCache for server flag checks
• Added proper error handling with safe fallbacks

Changes

New Files

• telemetry/exporter.go (192 lines) - Export infrastructure
• telemetry/exporter_test.go (448 lines) - Comprehensive exporter tests

Modified Files

• telemetry/config.go (+48 lines) - Added isTelemetryEnabled() function
• telemetry/config_test.go (+230 lines) - Added opt-in priority tests
• telemetry/DESIGN.md - Updated checklist (Phases 3-5 marked complete)

Test Coverage

All 70+ tests passing ✅

• ✅ 17 new exporter tests (success, retries, circuit breaker, tag filtering, error swallowing, exponential backoff, context cancellation)
• ✅ 8 new opt-in priority tests (all 5 priority levels, error handling, server scenarios)
• ✅ All existing tests continue to pass

Test run time: 2.017s

Testing Done

Unit Tests

• Export success scenarios with mock HTTP server
• Retry logic on 5xx errors with exponential backoff
• Non-retryable 4xx errors (no retry)
• 429 rate limiting (retryable)
• Circuit breaker integration (drops when open)
• Tag filtering (exports only allowed tags)
• Error swallowing (no panics)
• Context cancellation handling
• All 5 opt-in priority levels
• Server error scenarios
• Unreachable server handling

Integration Tests

• HTTP mock server tests for all scenarios
• Circuit breaker state transitions
• Feature flag cache integration

Design Alignment

This implementation follows the design document (telemetry/DESIGN.md) specifications:

• ✅ Section 3.6 - telemetryExporter
• ✅ Section 5 - Export Mechanism
• ✅ Section 6.4 - Opt-In Control & Priority

Related Issues

• Implements: PECOBLR-1143 (Phases 4-5)
• Depends on: PECOBLR-1143 Phases 1-3 (already merged)
• Enables: PECOBLR-1381 (Phase 6 - Collection & Aggregation)
• Enables: PECOBLR-1382 (Phase 7 - Driver Integration)

Checklist

• Implementation follows design document
• Comprehensive unit tests added
• All tests passing
• DESIGN.md checklist updated
• Code follows Go best practices
• Error handling with proper swallowing
• Thread-safe implementation
• No breaking changes

Next Steps

After this PR merges:

1. Phase 6 (PECOBLR-1381): Implement metric collection and aggregation
2. Phase 7 (PECOBLR-1382): Integrate with driver (connection.go, statement.go)

---

🤖 Generated with Claude Code

[samikshya-db]

Recreating with git stack for proper stacked PR management