fix cat.llm for wide-range compatibility

[lucagobbi]

Description

Recently different providers like Ollama, Anthropic, Google Gemini released some models that are not working correctly with cat.llm() method. The problem is caused by the direct usage of SystemMessage without a HumanMessage. The quick fix is to use HumanMessage instead of the SystemMessage, even if it could bring some problems due to different tokenizers as discussed in the latest dev meeting 2025/01/27.

Please try this with different models and providers
I tried with:

• Ollama llama3.2:3b
• Gemini gemini-2.0-flash-exp
• Anthropic claude-3-5-haiku-20241022
• OpenAI gpt-4o-mini

Related to issue #999 #1006 #1011

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist:

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas

[pieroit]

Thanks @lucagobbi
It's a temp solution but the best one