Add OpenTelemetry observability lab to Agent Memory Workshop

[markxnelson]

Summary

This PR extends the Agent Memory Workshop with a new Part 7 observability lab that makes the memory-aware agent inspectable with LangSmith.

It adds:

• A new Part 7 guide covering LangSmith setup and trace inspection
• Three new workshop TODOs for configuring telemetry, wrapping the agent loop with spans, and inspecting observed turns
• Student and complete notebook updates for the observed agent path
• Troubleshooting guidance for LangSmith, missing traces, and safe telemetry defaults

Why This Matters

The workshop already shows how Oracle AI Database gives agents durable memory and controlled context growth. This addition makes the runtime behavior visible.

Learners can now see the operational path behind an agent turn: memory reads, context assembly, tool lookup, LLM calls, context checks, tool execution, and memory writes. The lab keeps telemetry safe by recording counts, lengths, tool names, model names, and statuses rather than prompts, retrieved documents, or raw tool output.

Validation

• Checked current upstream contribution guidance in README.md
• Compared this branch against fetched upstream/main
• Validated all code with a full notebook execution and Docker Compose validation