Tracing landscape

[MatthewKhouzam]

Summary

Adds a new tracing_landscape/ subfolder importing the AAIF reference architecture collection — structured assessments of observability, tracing, and profiling technologies evaluated through the lens of AI agent infrastructure. The content is brought in with its full upstream commit history (via git subtree), so the 24 commits behind these documents are preserved rather than squashed.

What's included

The collection is organized bottom-up through the observability stack:

• 01 - Foundations: trace encoding and comparative context: CTF, LTTng/perf/FTrace comparison, Eclipse Trace Compass.
• 02 - Kernel Tracing: FTrace, Linux perf, LTTng (kernel + UST), and Linux driver tracing.
• 03 - Hardware Accelerators: GPU/AI tracing overview, NVIDIA Nsight, AMD roctracer/rocprofiler, PyTorch Profiler.
• 04 - Network: Wireshark deep packet inspection.
• 05 - AI Agent Observability: Agent Trace spec, Goose, Langfuse, PostHog, TMLL, OpenTelemetry, Datadog.

Every document assesses its subject across the five AAIF dimensions: Observability, Security, Identity, Reliability, and Accuracy.

Also added

• tracing_landscape/AGENTS.md - an agent-agnostic version of the aaif-ref-arch skill (Kiro-specific frontmatter removed), documenting the process, reference-architecture template, evaluation rubric, and writing style used to produce these assessments.
• The original .kiro/skills/aaif-ref-arch/SKILL.md is carried over as part of the imported history.

[mr-lee]

Wow this is a ton! Could we split out the skills from the other documents? I think we can handle them incrementally because otherwise this is a massive amount of stuff.