Move opencode config to static file; switch to vLLM-only backend

- Remove inline opencode_config_content() and configured_ollama_host()
  from Rexfile; config is now a plain file in opencode/opencode.json
- Remove JSON::PP dependency (no longer needed)
- Simplify home_opencode task to one-liner using ensure glob
- opencode/opencode.json: all 10 vLLM model presets from hyperstack
  (qwen3-coder-next, gpt-oss-20b/120b, llama-3.3-70b, qwen25-coder-32b,
  qwen3-coder-30b, deepseek-r1-32b, qwen3-32b, devstral, nemotron-super)
- Default model set to vLLM qwen3-coder-next (was Ollama qwen3-coder:30b)
- Ollama models removed; opencode now connects exclusively to vLLM

Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>

0 files changed, 0 insertions, 0 deletions