Fix #70: Best practice: HADS format for system prompts that work acro...

[JiwaniZakir]

Closes #70

Adds documentation for HADS (Human-AI Document Standard) as a best practice for writing portable system prompts across OpenRouter's 300+ supported models.

Changes

• docs/hads-system-prompts.md (new file): Explains the problem of system prompts breaking when switching model tiers, introduces HADS tags ([SPEC], [NOTE], [BUG]), includes a concrete API authentication example showing all three tag types, and links to the HADS specification.
• README.md (line ~44): Adds a "Portable System Prompts (HADS)" entry under the examples index, pointing to the new doc.

Motivation

System prompts that work on GPT-4 frequently degrade on smaller models (Llama 3 8B, Mistral 7B) because those models rely on structural inference rather than explicit instruction. HADS replaces implicit structure with direct semantic tags, producing consistent behavior across model tiers without any tooling — pure Markdown. This is directly relevant to OpenRouter users who switch models mid-project and need their system prompts to remain stable.

Testing

• Verified docs/hads-system-prompts.md renders correctly in GitHub's Markdown preview, including the fenced code block containing HADS-tagged content.
• Confirmed the README link [docs/hads-system-prompts.md](docs/hads-system-prompts.md) resolves correctly relative to the repo root.
• No runnable code was added; this is documentation only, so no test suite changes are required.

---

This PR was created with AI assistance (Claude). The changes were reviewed by quality gates and a critic model before submission.