Add OpenCode Zen and Go as separate model providers

[priyanshiranawat15]

Summary

• Add OpenCode Zen and OpenCode Go as separate providers with independent API keys, hosts, and curated model lists.
• Route each model to the wire OpenCode actually serves (Responses, Chat Completions, Anthropic Messages, or Gemini) instead of sending both products to one OpenAI-compatible endpoint.
• Send OpenWorker's user agent and a stable per-conversation x-opencode-session header, which OpenCode Go requires. Qwen and MiniMax on the Messages API do not get Claude-only options.

Fixes #109

Test plan

• Paste a Zen key and confirm only the Zen card shows connected; Go stays disconnected until its own key is saved
• Remove the Zen key and confirm the Go key remains
• Pick a Zen model and a Go model and confirm a turn completes
• pytest tests/test_opencode_provider.py

[priyanshiranawat15]

@rohitprasad15 / @andrewyng please review this pr, I have added support for open code go and zen.