feat: add support for llama models proper messaging

[jorgeantonio21]

Add support for Llama 2 models and update prompt formatting

Motivation

As Llama models continue to evolve, it's crucial to support both Llama 2 and Llama 3 models in our codebase. This PR aims to add support for Llama 2 models and update the prompt formatting logic to accommodate both model families, ensuring our application can work seamlessly with the latest advancements in language models.

Description

This PR introduces the following changes:

1. Added support for Llama 2 models:

   • Added new model variants: meta-llama/Meta-Llama-2-7b, meta-llama/Llama-2-7b-chat-hf, and meta-llama/Llama-2-70b-hf
   • Updated the Model enum to include these new variants

2. Implemented model-specific prompt formatting:

   • Added a new messages_to_prompt method to the Model enum
   • Implemented messages_to_llama2_prompt function for Llama 2 models
   • Implemented messages_to_llama3_prompt function for Llama 3 models
   • Updated the RequestBody::to_generate_request method to use the new model-specific prompt formatting

3. Updated the ToolCall struct:

   • Added a function_call_string method to generate a string representation of function calls for Llama 3 prompts

4. Added comprehensive unit tests:

   • Tests for both Llama 2 and Llama 3 prompt formatting
   • Tests for various message combinations and edge cases

5. Updated documentation and comments to reflect the new changes

The prompt formatting follows the official Llama documentation:

• Llama 2 Messaging formatting: docs
• Llama 3 Messaging formatting: docs

Breaking Changes

There are no breaking changes in this PR. The existing functionality for Llama 3 models remains unchanged, and the new Llama 2 support is added in a backwards-compatible manner.

However, users should be aware that the prompt formatting for Llama 2 models differs from Llama 3 models. When switching between model families, ensure that your input adheres to the correct format for the chosen model.