feat: support multiple LM Studio provider instances

[jleaders]

Summary

• Scans config.provider for all keys matching /^lm.?studio/i (e.g. lmstudio, lm-studio, lm-studio-wooden, lm-studio-alienware) and queries each in parallel for model discovery
• Previously only the exact key lmstudio was handled — lm-studio and any additional named instances were silently ignored
• Updates isLMStudioProvider type guard to match the same pattern so the chat.params hook correctly validates models on all instances

Changes

• enhance-config: extract processHost(), add findLMStudioProviders() to scan all lm?studio provider keys; run all hosts in parallel via Promise.allSettled
• config-hook: replace hardcoded lmstudio key check with isLMStudioProviderKey(); sum models across all providers for logging
• type-guards: update isLMStudioProvider to match /^lm.?studio/i
• test: add multi-provider test; fix mockClear → mockReset to prevent mock bleed-through between tests

Usage

Configure multiple LM Studio instances in opencode.json and the plugin will discover models from all of them automatically:

{
  "provider": {
    "lm-studio": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio (local)",
      "options": { "baseURL": "http://localhost:1234/v1" }
    },
    "lm-studio-wooden": {
      "npm": "@ai-sdk/openai-compatible",
      "name": "LM Studio (wooden)",
      "options": { "baseURL": "http://wooden.local:1234/v1" }
    }
  }
}

Test plan

• All 17 existing tests pass (npm test)
• Single provider config works as before
• Multiple lm-studio-* providers each populate with models from their respective hosts
• Offline hosts are skipped gracefully without blocking other providers

🤖 Generated with Claude Code