Fixed bad interaction lines that were added in Ollama tool calling

[ahyatt]

Also improve llm-chat-prompt-to-text for betting prompt debugging.

This should resolve the issue noted in #224 (comment)

[s-kostyaev]

For now it looks like it doesn't work for me. Couple of smart enough small language models for testing:

• Ministral 3 3b https://ollama.com/library/ministral-3
• Qwen3 4b 2507 both thinking and instruct https://ollama.com/library/qwen3

[s-kostyaev]

It works now! Thank you!