[codex] Fix OpenClaw gateway v4 and emoji display handling

[ndrco]

What changed

Updated Listener for the latest OpenClaw Gateway protocol and kept emoji-display working even when speaker playback is disabled.

Why

The recent OpenClaw update moved Gateway clients to protocol v4. Listener still used protocol v3 in both its input-side and speaker-side Gateway clients, which caused websocket compatibility failures. While fixing that, we also decoupled emoji-display delivery from TTS playback so emoji still show when speaker is turned off.

Impact

• Listener websocket calls now negotiate protocol v4.
• SpeakerAgent can stay connected for emoji delivery when speech is disabled.
• Listener systemd template exports a PATH that includes the local openclaw binary for CLI fallback.

Validation

• /home/re/Applications/Listener/.venv/bin/python -m pytest /home/re/Applications/Listener/tests/test_openclaw_gateway.py
• /home/re/Applications/Listener/.venv/bin/python -m pytest /home/re/Applications/Listener/tests/test_speaker_agent.py