feat: add Avian as a named LLM provider

[avianion]

Summary

• Add Avian as a named OpenAI-compatible LLM provider
• API base: https://api.avian.io/v1, auth via AVIAN_API_KEY env var
• Models include deepseek/deepseek-v3.2 (164K ctx), moonshotai/kimi-k2.5 (131K ctx), z-ai/glm-5 (131K ctx), minimax/minimax-m2.5 (1M ctx)
• Supports chat completions, streaming, and function calling/tools

Changes

• pkg/config/config.go — Add Avian field to ProvidersConfig, update IsEmpty() and HasProvidersConfig()
• pkg/config/defaults.go — Add Avian model entries to default ModelList
• pkg/config/migration.go — Add Avian migration rule for legacy config format
• pkg/config/migration_test.go — Update AllProviders test count (18 → 19)
• pkg/providers/factory.go — Add avian case to explicit and inferred provider selection
• pkg/providers/factory_provider.go — Add avian to OpenAI-compatible protocol list and getDefaultAPIBase()
• config/config.example.json — Add Avian provider entry

Usage

{
  "model_list": [
    {
      "model_name": "deepseek-v3.2",
      "model": "avian/deepseek/deepseek-v3.2",
      "api_key": "your-avian-api-key",
      "api_base": "https://api.avian.io/v1"
    }
  ]
}

Or via legacy provider config:

{
  "providers": {
    "avian": {
      "api_key": "your-avian-api-key"
    }
  }
}

Test plan

• Existing tests pass (go test ./pkg/config/... ./pkg/providers/...)
• TestConvertProvidersToModelList_AllProviders updated for 19 providers
• Verified all changes follow existing provider patterns (Mistral, DeepSeek, Cerebras)

cc @yinwm @mengzhuo @alexhoshina — would appreciate a review when you get a chance. This follows the same pattern as previous provider additions (#333, #167, #541).

[avianion]

Hey @yinwm, would love your review on this when you get a chance. Happy to address any feedback!

[alexhoshina]

LGTM

[yinwm]

@avianion hi, please fix tests
LGTM

[Orgmar]

@avianion Cool to see Avian added as a provider! The model lineup with deepseek-v3.2, kimi-k2.5, and minimax-m2.5 gives users some interesting options, especially that 1M context window on minimax.

We've got the PicoClaw Dev Group on Discord for contributors to collaborate. If you're interested, email support@sipeed.com with the subject [Join PicoClaw Dev Group] avianion and we'll send you the invite link!