feat(mcp): add openai_generic provider for LiteLLM/Ollama compatibility

[lehcode]

Summary

Add support for openai_generic LLM provider in the MCP server factory.

This provider uses OpenAIGenericClient which calls /chat/completions with response_format for structured output instead of the /responses endpoint.

Problem

The default openai provider uses OpenAIClient which calls OpenAI's /responses endpoint. This endpoint is only available on OpenAI's native API and fails with:

• LiteLLM proxy
• Ollama
• vLLM
• Any other OpenAI-compatible API

When using these backends, users get errors like:

'NoneType' object has no attribute 'encode'

Solution

Add openai_generic case to LLMClientFactory that:

1. Imports OpenAIGenericClient (which already exists in graphiti-core)
2. Uses the same OpenAI provider config (api_key, api_url)
3. Routes to /chat/completions endpoint with response_format for structured output
4. Implements automatic fallback from json_schema to json_object for providers that don't support strict schemas
5. Adds robust JSON extraction for responses with trailing explanatory text

Usage

llm:
  provider: "openai_generic"  # Instead of "openai"
  model: "your-model"
  providers:
    openai:
      api_key: ${OPENAI_API_KEY}
      api_url: ${OPENAI_BASE_URL}  # e.g., http://localhost:4000

---

Type of Change

• Bug fix (JSON extraction, auto-fallback)
• New feature (openai_generic provider)
• Performance improvement
• Documentation/Tests

Objective

Enable Graphiti to work with OpenAI-compatible APIs like LiteLLM proxy and Ollama. Some providers (e.g., Gemini via LiteLLM) don't support json_schema response format and return the schema definition instead of data. This PR adds automatic detection and fallback to json_object mode, plus robust JSON extraction for providers that append explanatory text after JSON output.

Testing

• Unit tests added/updated (18 tests for _is_schema_returned_as_data, _extract_json)
• Integration tests added/updated
• All existing tests pass
• Verified with LiteLLM proxy + Ollama backend
• Confirmed /chat/completions endpoint is called (not /responses)
• Tested add_memory and entity extraction working

Breaking Changes

• This PR contains breaking changes

Checklist

• Code follows project style guidelines (make lint passes)
• Self-review completed
• Documentation updated where necessary
• No secrets or sensitive information committed

Related Issues

N/A

[danielchalef]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[lehcode]

I have read the CLA Document and I hereby sign the CLA

[lehcode]

Added auto-fallback for providers without json_schema support: e680665