Add a way to opt out of thinking mode for LLMs

[chmjkb]

Problem description

Currently, to disable thinking mode for models such as Qwen, users must know to prefix their prompt with the token, which is not ideal from a UX perspective.

Proposed solution

Add a prop to disable thinking mode (e.g., disableThinking or thinkingMode: false)

Alternative solutions

No response

Benefits to React Native ExecuTorch

This will provide a cleaner API for users who want to opt out of thinking mode, eliminating the need for manual token management.

Additional context

No response