feat: add Z.AI coding plan support

[dpshade]

Hey 👋 , thanks for creating and maintaining this minimalistic LLM tool!

This pull request adds support for Z.AI's coding plan models while maintaining the tool's lightweight ethos and original defaults. Users can now use GLM models optimized for coding tasks.

Changes Made

• Add Z.AI model shortcuts: -a for GLM-4.5-air (fastest), -z for GLM-4.6 (advanced)
• Preserve original default: Keeps qwen/qwen3-235b-a22b-2507:nitro as default
• Auto-detect Z.AI endpoint: Uses ZAI_BASE_URL env var, falls back to OpenRouter
• Zero breaking changes: Full backward compatibility maintained

Z.AI Usage

# Configure Z.AI (once)
export ZAI_BASE_URL="https://api.z.ai/api/coding/paas/v4"
export OPENROUTER_API_KEY="your-zai-api-key"

# Use Z.AI models
ask -a "Write a Python function"        # Uses GLM-4.5-air (fast)
ask -z "Explain this algorithm"         # Uses GLM-4.6 (advanced)

# Original behavior unchanged
ask "Quick question"                    # Still uses qwen by default

OpenRouter Compatibility

Original functionality is completely preserved - when ZAI_BASE_URL is not set, the tool works exactly as before with all original models and defaults.

Why This Change

Z.AI's coding plan offers excellent performance for coding tasks, and GLM-4.5-air is particularly fast for quick command generation and code snippets. The implementation adds only 12 lines while maintaining the minimalistic approach and zero breaking changes.