Compatibility Implementation for Ragas.io with WatsonxLLM Parameters Handling

[JulioSanchezD]

Issue:
Ragas.io uses the LangchainWrapper to leverage an LLM for its evaluation tasks. However, when calling the invoke method, it adds text generation parameters outside of the ‘params’ dictionary, which necessitates modifications to ensure compatibility and proper parameter handling.

Added Methods:

1. _validate_chat_params: Validates and adjusts the ‘temperature’ parameter in the ‘params’ dictionary to ensure it meets the required conditions. Only the ‘temperature’ parameter is currently validated. More cases can be added as needed.
2. _override_chat_params: Overrides the class parameters with those provided in the invoke method. Updates ‘params’ with matching keys from kwargs and removes those keys from kwargs. It includes the merge from the params dictionary in the invoke method too. This ensures that the parameters sent from ragas are correctly integrated into and handled WatsonxLLM within the existing parameter structure.
3. test_watsonxllm_invoke_with_params_4: Implemented the CI test for this implementation

[MateuszOssGit]

@JulioSanchezD I have made some changes to be backward compatibility.
Also there is no need to check the value of the "temperature" parameter. It is validated by the backend

[Mateusz-Switala]

LGTM