feat(nvidia): replace GLM-5.1 with GLM-5.2

[stupidloud]

Summary

• NVIDIA's NIM catalog (integrate.api.nvidia.com/v1/models) no longer lists z-ai/glm-5.1; it now serves z-ai/glm-5.2 in its place.
• Renamed providers/nvidia/models/z-ai/glm-5.1.toml to glm-5.2.toml and updated name/description/dates, dropped the deprecated status, bumped context to 1,000,000, and added a sourced comment for the chat_template_kwargs.enable_thinking reasoning toggle.

Closes #3002

Sources:

• https://build.nvidia.com/z-ai/glm-5.2
• https://docs.api.nvidia.com/nim/reference/z-ai-glm-5.2
• curl https://integrate.api.nvidia.com/v1/models (confirms z-ai/glm-5.1 absent, z-ai/glm-5.2 present)

Test plan

• Parsed the TOML with tomllib to confirm it's well-formed and has all required schema fields (no bun available in this environment to run bun validate)