feat: add openbrowser skill for AI-driven testing workflows

[digi4care]

Summary

This PR adds a complete openbrowser skill that helps AI coding agents use open-browser as a lightweight alternative to Playwright for semantic testing and browsing workflows.

What is a "skill"?

A skill is a structured knowledge pack that AI agents (like Pi, Cursor, Copilot, Claude, etc.) can load to understand when and how to use a tool. It provides decision matrices, workflow patterns, and reference documentation — so the AI agent makes good choices without guessing.

What's included

File	Content
skills/openbrowser/SKILL.md	Main reference — when to use, when NOT to use, command matrix, 6 workflow patterns, error handling
skills/openbrowser/references/cli-reference.md	All 7 subcommands with flags and examples
skills/openbrowser/references/testing-patterns.md	Playwright → open-browser translation table, assertion recipes
skills/openbrowser/references/examples.md	10 end-to-end examples (login flows, form testing, site mapping, etc.)
skills/openbrowser/references/repl-and-cdp.md	REPL commands and CDP WebSocket server for automation
skills/openbrowser/references/knowledge-graph.md	Site mapping, BFS crawler, state fingerprinting
skills/openbrowser/references/semantic-roles.md	HTML element → ARIA role/action mapping table
skills/openbrowser/references/programmatic-usage.md	Rust Browser API documentation
skills/openbrowser/references/installation.md	Build from source, Arch Linux specifics, Docker
skills/openbrowser/references/troubleshooting.md	Build/runtime issues with fixes
skills/openbrowser/references/registry.json	Machine-readable index of all references

Key design decisions

1. Decision matrices over feature lists — the skill tells AI agents when to pick open-browser vs Playwright, not just what open-browser does
2. 6 workflow patterns — page verification, form interaction, navigation flows, site mapping, network debugging, session persistence
3. Playwright translation table — helps agents migrate existing Playwright test patterns to open-browser equivalents
4. Honest limitations section — clearly documents what open-browser cannot do (no JS rendering, no screenshots, no cross-browser)

Testing

• SKILL.md audit score: 101/100 PASS
• All CLI examples tested against current open-browser CLI
• Arch Linux build requirements verified (cmake, clang, LIBCLANG_PATH)

Why this is useful for the project

Users who discover open-browser through AI coding tools (increasingly common) will have immediate access to structured guidance. The skill makes it trivial for any AI agent to:

• Understand open-browser's capabilities and limitations
• Choose the right subcommand for the task
• Follow proven workflow patterns
• Troubleshoot common build and runtime issues

This lowers the adoption barrier significantly for AI-assisted development workflows.