feat: add Qwen3.8 Flash Next renderer

[bclavie]

Add Qwen/Qwen3.8-Flash-Next model to renderer and multimodal maps

• Registers Qwen/Qwen3.8-Flash-Next in MODEL_RENDERER_MAP to resolve to the qwen3.8 renderer and marks it image-capable in MULTIMODAL_MODELS in base.py
• Sets _ENABLE_THINKING_DEFAULTS for Qwen/Qwen3.8-Flash-Next to True in qwen35.py, so thinking is on by default when enable_thinking is unset
• Extends test matrices across bridge, roundtrip, renderer config parity, disabled-thinking stability, and Qwen3.8-specific tests to cover the new model; refactors test_qwen38.py to parametrize over a shared QWEN38_MODELS list

^{Macroscope summarized 6e5e61b.}