docs: add guide for using Headroom with OpenCode Go

[suvenkatesh97]

Documents how to configure Headroom proxy with OpenCode Go for OpenCode users.

• Two-proxy setup compresses all 22+ Go models (OpenAI + Anthropic format)
• Includes official pricing & limits tables from opencode.ai/docs/go
• Model-to-proxy mapping reference for both formats
• Covers Kimi K3, Grok 4.5, GLM 5.2, MiniMax M3, Qwen3.7, DeepSeek V4 (via Go), and more
• Cross-references open-code-deepseek.md for direct DeepSeek API setup
• All API keys use placeholders

Description

Adds documentation (docs/open-code-go.md) showing OpenCode users how to route Go's subscription models through Headroom proxy for input compression and output shaping. OpenCode Go ($5 first month, then $10/month) provides 22+ open-source coding models — this guide enables compression across ALL of them.

Type of Change

• Documentation update

Changes Made

• New file: docs/open-code-go.md — complete setup guide covering Go subscription, two-proxy architecture, OpenCode provider config, model-to-proxy mapping, pricing tables, and troubleshooting

Testing

• Unit tests pass (pytest)
• Linting passes (ruff check .)
• Type checking passes (mypy headroom)
• New tests added for new functionality
• Manual testing performed

Test Output

N/A — docs-only change, no code paths touched. No pytest/ruff/mypy relevant.

Real Behavior Proof

• Environment: Ubuntu, Python 3.13, headroom-ai 0.32.1, OpenCode (Go CLI)
• Exact command / steps: Started two Headroom proxy instances (ports 8788, 8789) routing to OpenCode Go API, configured OpenCode with headroom-go (OpenAI) and headroom-go-anthropic (Anthropic) providers, tested chat completions through both proxies using Go API key via curl.
• Observed result: kimi-k3 returned "Hello" (83 completion tokens, 67 reasoning tokens, 90 cached). grok-4.5 returned "Hi there! How can I help you today". glm-5.2 returned "Hello! How can I help you today?" (124 completion tokens with reasoning). hy3 returned "Hello! How can I help you today?" with reasoning content. minimax-m3 through Anthropic proxy returned "Hello! How can I help you today?" (111 input, 9 output, 114 cached). SmartCrusher compression active on both proxies — 13008 tokens saved on port 8788. Output shaping level 2 active on all instances.
• Not tested: did not manually exercise the full OpenCode TUI flow end-to-end against a real Go subscription with real prompts (used raw HTTP requests via curl to verify proxy routing and compression).

Review Readiness

• I have performed a self-review
• This PR is ready for human review

[github-actions]

PR governance

This PR follows the template and is marked ready for human review.