Use Codex fast service tier

[MagMueller]

Summary

• model, reasoning effort, and fast mode are now separate Codex settings
• /fast and plain fast toggle service_tier="priority" instead of changing reasoning effort
• /model shows fast mode separately from low/medium/high/xhigh and can toggle it inline

Verification

• confirmed local Codex model catalog exposes GPT-5.5/GPT-5.4 fast as service_tiers: priority/Fast
• confirmed this lane was actually invoked with -m gpt-5.5 -c model_reasoning_effort="xhigh"

Tests

• /home/bux/.local/bin/ruff check agent/telegram_bot.py agent/test_telegram_bot.py
• /opt/bux/venv/bin/python -m unittest agent/test_telegram_bot.py

---

Summary by cubic

Switches Codex “fast mode” to use the fast service tier instead of lowering reasoning effort, and separates Codex settings into model, reasoning effort, and fast mode. This gives clearer controls and correctly invokes the priority tier.

• New Features
  • Added service_tier to per-lane Codex settings and pass -c service_tier="priority" when enabled.
  • /fast and plain fast now toggle fast mode on/off without changing model_reasoning_effort.
  • /model picker shows “Fast mode: on/off” and supports inline toggle; effort labels use simple titles.
  • Updated tests to cover service tier toggling, UI state, and settings persistence.

^{Written for commit 36ff2ad. Summary will update on new commits. Review in cubic}

[cubic-dev-ai]

No issues found across 2 files

_{Re-trigger cubic}