feat: OpenAI compatibility

[andthattoo]

Implement comprehensive test suite to ensure 100% compatibility with OpenAI API specification for dnet.

• Validate all OpenAI API endpoints
• Test request/response format compatibility
• Verify error handling matches OpenAI behavior
• Ensure streaming responses work correctly

Test Coverage Needed

• /v1/chat/completions endpoint
  • Standard requests
  • Streaming responses
  • Function calling
  • All supported parameters (temperature, max_tokens, etc.)
• /v1/embeddings endpoint
• /v1/models endpoint
• Authentication headers: the user will just be prompted to use dnet as the API key, although the API server will ignore the bearer token regardless (similar to Ollama)
• Error response formats
• Edge cases (empty requests, invalid parameters)

Acceptance Criteria

• All tests pass against both OpenAI API and our implementation
• Error codes and messages align with OpenAI standards

[erhant]

Will create a tests/openai_compat.py integration test for this.

[erhant]

See https://github.com/ollama/ollama/blob/main/docs/openai.md for reference of a non-complete & sufficient compatibility.

[erhant]

@andthattoo will add /v1/models/list endpoint to the API, which will return the list of models (repo ids) supported by dnet. This will also be used by TUI to show the user the model selection menu.