feat: enhance LLM routing and support for native implementations

- Implemented provider prefix routing for LLMs, allowing for native implementations of OpenAI, Anthropic, and Google Gemini.
- Added fallback to LiteLLM if native implementations are unavailable.
- Updated create_llm function to support provider prefixes and native preference settings.
- Introduced new LLM classes for Claude and Gemini, ensuring compatibility with existing LLM interfaces.
- Enhanced documentation for LLM utility functions to clarify usage and examples.

src/crewai/llms/__init__.py

@@ -1 +1,11 @@
-"""LLM implementations for crewAI."""
+"""CrewAI LLM implementations."""
+
+from .base_llm import BaseLLM
+from .openai import OpenAILLM
+from .anthropic import ClaudeLLM
+from .google import GeminiLLM
+
+# Import the main LLM class for backward compatibility
+
+
+__all__ = ["BaseLLM", "OpenAILLM", "ClaudeLLM", "GeminiLLM"]