Observability 2.0: Tracing AI "Thought Chains" with OpenTelemetry

apcore presents an observability pattern for agentic AI systems using OpenTelemetry spans around each `Executor.call()`.

It adds AI-specific metadata, propagates trace context across services, and exposes metrics for counts, latency, and errors.

Tracing can be enabled by registering `TracingMiddleware`, making agent behavior easier to inspect in production.

The approach helps teams debug multi-step reasoning and tool use, where failures often lack a clear stack trace.