Restructure agent context for cross-agent use (AGENTS.md + .agents/) (#32)

Make agent-neutral files canonical so Claude Code, Codex, and other
agents share one set of instructions with no duplication:

- CLAUDE.md (root, client/, server/) -> AGENTS.md, with CLAUDE.md kept
  as a symlink (Claude Code reads CLAUDE.md; Codex/Cursor/etc. read
  AGENTS.md, including nested ones)
- .claude/rules/ -> .agents/rules/, with .claude/rules left as a dir
  symlink (Claude Code resolves symlinked rules; other agents are
  pointed at the files from AGENTS.md). The how-to-write-rules
  CLAUDE.md becomes .agents/rules/README.md
- agent-browser skill -> .agents/skills/ next to shadcn, with per-skill
  symlinks in .claude/skills/ — the same layout the skills CLI already
  used for shadcn. Codex discovers .agents/skills/ natively
- Document the layout in AGENTS.md and update path references in
  DESIGN.md and docs/browser-testing-cloud.md


Claude-Session: https://claude.ai/code/session_016LEyuB1y6D84r4RSNkzXRW

Co-authored-by: Claude <noreply@anthropic.com>

Author: tonyfresher