fix: remove reasoning toggles for GLM-4.5 and GLM-4.6 on z.ai provider

[roomote]

Removes the reasoning toggle UI for GLM-4.5 and GLM-4.6 models on the z.ai provider, as confirmed by the z.ai team that these models were trained without think/reasoning data for coding agents.

Changes Made

• Removed supportsReasoningBinary: true from GLM-4.5 (both international and mainland variants)
• Removed supportsReasoningBinary: true from GLM-4.6 (both international and mainland variants)
• Removed all reasoning-related tests from zai.spec.ts since no current z.ai models support reasoning

Context

According to the z.ai team: "For scenarios in coding agents, GLM-4.5 and 4.6 were trained with non-think data. We might add think in the next-generation, but it will be a new model. Think in coding agents will be available in GLM-4.6 Air."

This means the reasoning toggle should not be shown for these models, as they do not support extended thinking/reasoning capabilities in coding agent scenarios.

Testing

• ✅ All zai.spec.ts tests pass (19 tests)
• ✅ Linting passes
• ✅ Type checking passes

---

View task on Roo Code Cloud

---

Important

Remove reasoning toggle for GLM-4.5 and GLM-4.6 models on z.ai provider due to lack of reasoning data support.

• Behavior:
  • Removed supportsReasoningBinary: true from GLM-4.5 and GLM-4.6 models in zai.ts.
  • Removed reasoning-related tests from zai.spec.ts as no z.ai models support reasoning.
• Context:
  • GLM-4.5 and 4.6 models were trained without reasoning data for coding agents, confirmed by z.ai team.
• Testing:
  • All tests in zai.spec.ts pass after changes.

^{This description was created by for 16c3e10. You can customize this summary. It will automatically update as commits are pushed.}

[roomote]

Rooviewer   See task on Roo Cloud

Review complete. No issues found.

The changes correctly remove supportsReasoningBinary from GLM-4.5 and GLM-4.6 models (both international and mainland variants) and clean up the corresponding tests. This aligns with the z.ai team's guidance that these models were trained without think/reasoning data for coding agents.

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}

[roomote]

Rooviewer   See task on Roo Cloud

Review complete. No issues found.

The changes correctly remove supportsReasoningBinary from GLM-4.5 and GLM-4.6 models (both international and mainland variants) and clean up the corresponding tests. This aligns with the z.ai team's guidance that these models were trained without think/reasoning data for coding agents.

_{Mention @roomote in a comment to request specific changes to this pull request or fix all unresolved issues.}