feat(cli): add openai provider support for OpenAI-compatible endpoints

[Thump604]

Summary

Adds the openai provider to the CLI's supported providers list, enabling use of OpenAI-compatible API endpoints. This allows the CLI to connect to locally deployed LLMs (vLLM, llama.cpp, Ollama, LM Studio, etc.) and any OpenAI-compatible inference server.

• Add "openai" to supportedProviders in types.ts
• Add openai: "OPENAI_API_KEY" to envVarMap in provider.ts
• Add "openai" case in getProviderSettings() with OPENAI_BASE_URL env var support
• Update README environment variable table

The openai provider already exists in @roo-code/types as a customProvider with full schema support (openAiBaseUrl, openAiApiKey, openAiModelId), and the core extension already handles it via OpenAiHandler. The CLI simply needed to expose it in its provider whitelist.

Usage

export OPENAI_API_KEY=sk-local
export OPENAI_BASE_URL=http://localhost:8080/v1
roo --provider openai --model my-model "Hello"

Or via settings file (~/.roo/cli-settings.json):

{
  "provider": "openai",
  "model": "my-model"
}

Difference from openai-native

	openai	openai-native
API	Chat Completions (/v1/chat/completions)	Responses API
Custom base URL	Yes, via OPENAI_BASE_URL	Limited
Local LLMs	Yes (vLLM, llama.cpp, Ollama, etc.)	No
Settings fields	openAiBaseUrl, openAiApiKey, openAiModelId	openAiNativeApiKey

Test plan

• Verified locally with vLLM-MLX serving Qwen 3.5 122B on http://127.0.0.1:8080/v1
• Confirmed --provider openai --model <model> -k <key> resolves correctly
• Confirmed OPENAI_API_KEY and OPENAI_BASE_URL env vars are read
• Type-checked: "openai" satisfies ProviderName (already in customProviders)

Closes #11917

Interactively review PR in Roo Code Cloud