genai-function-calling: update to qwen3 model which reduces edge cases

[codefromthecrypt]

This updates to qwen3 model which reduces edge cases as all except spring-ai can use qwen3:0.6b regardless of if the major_version is needed by the LLM when answering the question.

Notes:

• a follow-up PR should probably happen to remove the think tags from reasoning models (qwen3 is reasoning) Here's an example https://github.com/vercel/ai/blob/main/content/docs/03-ai-sdk-core/40-middleware.mdx#extract-reasoning Another option is to disable thinking mode, but that's not yet implemented in ollama.
• That leaving out major_version requires a larger model in spring-ai is likely due to how the tool requests are translated in to a chat completion schema. I've not looked carefully at why and it is possible a tweak could be made to spring-ai to prevent needing a larger model here. This edge case is not new and also qwen2.5:3b was needed before while now we can get by with qwen3:1.7b