[FEATURE] OpenTelemetry integration — structured tracing, metrics, and logging

[ElioNeto]

Description

ApexStore has basic Prometheus metrics (#149) but no distributed tracing or correlation IDs. Production systems require OpenTelemetry for debugging latency, identifying bottlenecks, and correlating operations across services.

Proposed Implementation

1. Add opentelemetry crate dependency
2. Instrument all engine operations with tracing spans (set, get, delete, scan, flush, compact)
3. Propagate trace context from HTTP layer through to storage engine
4. Export traces via OTLP (Jaeger, Tempo, Datadog)
5. Add per-operation latency histograms with configurable buckets
6. Expose trace_id in HTTP response headers for debugging

Impact

• Required for production SRE / observability
• Enables latency breakdown (WAL fsync vs memtable vs compaction)
• Critical for debugging performance issues in production

Labels

[ElioNeto]

Fixed in commit 1d8c830.