v0.3.1

UX completion patch for v0.3.0's knowledge-service positioning. Removes
the "go clone GitHub and copy this file" friction for IDE / agent
integration. Also revamps top-level --help to surface all current
capabilities (audit / knowledge / integration) at a glance.

Added

• a11y-moda init <ide> subcommand — one-line install of bundled
  IDE / agent integration templates. No more "open GitHub → find
  examples/ → copy → paste".
  • init claude-code → ~/.claude/skills/a11y-moda/ (full skill dir
    with SKILL.md + REFERENCE.md)
  • init cursor → ./.cursorrules
  • init copilot → ./.github/copilot-instructions.md
  • init aider → ./.aider.conf.yml
  • init agent → stdout (or --dest for file write); platform-agnostic
    AGENT.md for any LLM agent (Cline, Continue, RooCode, custom)
  • init --list — list all IDEs + their default install paths
  • init <ide> --print — preview content (don't write); useful for CI
    / piping to other commands / pasting into agent prompts
  • init <ide> --dest <path> — override default install path
  • init <ide> --force — overwrite existing destination files
• Bundled examples in package — _examples/ directory now ships
  with the wheel via [tool.setuptools.package-data]. End users no
  longer need to clone the repo.
• Comprehensive --help — top-level help text now lists all
  capabilities organised into AUDIT / KNOWLEDGE / INTEGRATION /
  INSTALL sections. Single screen overview for new users.

Changed

• examples/ → src/a11y_moda/_examples/ — moved into the package
  so it ships with pip install a11y-moda. GitHub browse paths updated
  in README, README.en, docs/AI_INTEGRATION.md. Old examples/ URLs in
  CHANGELOG entries 0.3.0 left as historical record.
• docs/AI_INTEGRATION.md §11 — recommends a11y-moda init <ide>
  as the primary install path; manual copy from GitHub demoted to
  fallback.
• SKILL.md README — install instructions simplified from
  "git clone + curl" to "a11y-moda init claude-code".

Notes

• Why no interactive prompt — a11y-moda init (no IDE arg) prints
  ERROR + clear options list rather than a click.prompt loop. AI
  agents (Claude Code, Cursor, Copilot, Aider) shell out without TTY
  stdin and would hang on interactive prompts. Single-line
  init <ide> works the same for AI and human users.
• Why --force required for overwrite — .cursorrules,
  SKILL.md, etc. are commonly hand-edited. Silent overwrite would
  destroy user work. Default behavior: refuse with helpful ERROR;
  --force opt-in to replace.
• Why init agent has no default file path — the AGENT.md content
  is meant for the agent's system prompt setting (location varies
  by agent: Cline = workspace settings, Continue = config.json,
  custom = code). Default = stdout for piping; --dest for file.
• AI install flow now possible: user says "set up a11y-moda for
  Cursor", agent runs pip install a11y-moda && a11y-moda init cursor,
  done. No human navigation through GitHub UI.