Gemini 3: Add “Thinking Level” support (send thinking config in request)

Description:
TypingMind currently supports using Gemini 3 for chat, but it does not provide or apply a Thinking Level configuration.

What I observed

• Using Charles Proxy, the outgoing request to Gemini 3 does not include thinking-related configuration fields when using TypingMind.

• As a result, Gemini 3 runs with default thinking behavior.

What I expect

• TypingMind should expose a “Thinking Level” (or equivalent) control for Gemini 3.

• When set, TypingMind should include the correct thinking configuration in the request payload.

Why this matters

• Thinking Level materially affects response quality/cost/latency, and is important for consistent results across tools.

Evidence

• Charles Proxy screenshots showing the request payload (sanitized) can be provided.

•