Configure AI Assistant output limits and reliably turn off Ollama thinking

Allow developers to configure the maximum response tokens in Settings > AI Assistant, so models have enough room to complete code conversions and other substantial tasks. Explain that the response allowance is separate from the model's context window and may also be consumed by thinking.

Ensure that turning off "Enable thinking" explicitly requests that Ollama disable thinking, rather than leaving the choice to the model's default. Models that cannot disable thinking should report their limitation clearly.

These changes should reduce unexpectedly truncated responses and make the assistant's generation settings behave predictably.