Fix/daemon down ux

[RonCodes88]

Summary

Fixes the red PreToolUse:hook error / Stop hook error banners that showed up across Claude Code, Codex, and Cursor when the OpenAgentLock daemon was off (or when the published CLI
tarball was missing the wrapper script). Adds a live OpenAgentLock ✓ / ⚠ daemon offline indicator under the Claude Code chat via statusLine. Codex and Cursor get silent fail-open
instead of a chat-injected nudge — anything we put in those harness chat fields registered as a prompt-injection attempt.

What changed

• New stable wrapper at <agentlockHome>/bin/agentlock written by agentlock install; hook entries reference it instead of the volatile node_modules path, so package upgrades can't
  strand the wired command.
• Shell-quote the binary path in every harness's command string (Claude / Codex / Cursor) so paths with spaces (Library/Application Support) survive /bin/sh -c parsing.
• Claude Code statusLine config wired automatically by the installer to a tiny health-check script (<binDir>/agentlock-status); shows OpenAgentLock ✓ when the daemon is up,
  OpenAgentLock ⚠ daemon offline when it's not. Output is pure UI — never enters the model's input stream.
• All three shims now silent fail-open on transport errors (no stdout, no stderr, exit 0). Removed the additionalContext / agent_message chat-injection paths.
• prepublishOnly guard in cli/package.json asserts the source-tree wrapper is in the tarball before npm publish — catches the kind of regression that produced 0.1.12.
• Bumped to 0.1.15.

Checklist

• I read CONTRIBUTING.md
• No secrets, real tokens, or .env content committed
• Public docs updated if user-facing behavior changed