docs(harness): HARNESS-PHILOSOPHY.md — why/what/how/when with limitations

[magicpro97]

User Story

As a developer onboarding to harness engineering with sk, I want a clear philosophy document that explains why harness engineering matters, what it does, how it works, and its limitations, so that I can make informed decisions about when and how to use it in my projects.

Acceptance Criteria

• docs/HARNESS-PHILOSOPHY.md created with full content
• Document covers: Why (the +36% performance gap), What (components), How (sk architecture), When (decision matrix), Architecture, Limitations, Best Practices, Further Reading
• Includes concrete examples and references to SWE-bench, AgentBench, CORE benchmark papers
• Mentions limitations: non-determinism, cost, latency, multi-agent races, scope of SK_HARNESS
• Cross-referenced from docs/HARNESS.md (update cross-reference)
• Linked from HARNESS-PHILOSOPHY.md to docs/HARNESS.md for API details

Priority: HIGH

Labels: documentation, harness, priority:high

Files: docs/HARNESS-PHILOSOPHY.md, docs/HARNESS.md

[magicpro97]

Implemented in PR #734. docs/HARNESS-PHILOSOPHY.md (552 lines) covers full why/what/how/when/limitations. Cross-referenced from docs/HARNESS.md.