Add Custom Generation function support

[Pevooo]

✨ Add CustomTextGenerator for Custom Function-Based Text Generation

📌 Summary

This PR introduces CustomTextGenerator, a new text generation class for llmx that allows users to define their own text generation function. This enables custom logic for generating text, making llmx more flexible for various use cases.

🛠️ Changes Introduced

• New CustomTextGenerator Class
  • Accepts a text_generation_function: Callable[[str], str] to generate text based on a user-defined function.
  • Implements caching to reuse previously generated responses.
  • Includes token counting functionality via num_tokens_from_messages().

✅ How It Works

# Define a custom text generation function
def my_custom_function(prompt: str) -> str:
    return f"Generated response for: {prompt}"

# Instantiate CustomTextGenerator
custom_gen = CustomTextGenerator(text_generation_function=my_custom_function)

# Generate text based on a user message
response = custom_gen.generate([{"role": "user", "content": "Hello"}])
print(response.text[0].content)  # Output: "Generated response for: user: Hello"

🚀 Why This is Useful

• Enables custom text generation logic beyond built-in providers.
• Improves flexibility for users integrating external models or APIs.
• Supports caching, reducing redundant calls and improving efficiency.