feat(ai): switch chat model to qwen3.8-flash (#325)

[fworks-tech]

Description

The chat widget's hard-coded deepseek-v4-flash is failing at the OpenCode gateway (404 inference_failed on /inference/openai/v1/chat/completions — workspace log e6eede50-1df7-4336-a658-374572f6ebc0, while the same key returns 200s for MiMo-V2.6-Flash). The model is still listed in /zen/v1/models, so this is the deepseek route breaking upstream, not our key or code.

Switches MODEL_ID to qwen3.8-flash (official Zen docs: @ai-sdk/openai-compatible at the same /zen/v1/chat/completions endpoint, $0.15 in / $0.47 out per 1M — verified LISTED live with the production key) and moves TOKEN_COST_PER_1M to 0.47 so the $0.50/h abuse budget keeps charging honestly. cost.test.ts expectations updated.

Fixes #325

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that causes existing functionality to not work as expected)
• Documentation update
• Refactoring (no functional changes)

How Has This Been Tested?

• Tested locally — npm run lint, npm run typecheck clean; npm test 734/734 across 103 files (incl. updated cost.test.ts)
• All existing tests pass
• New tests added (if applicable) — expectation update covers the new rate

Checklist

• My code follows the project's style guidelines
• I have performed a self-review of my code
• I have commented my code where necessary
• I have updated documentation accordingly
• My changes generate no new warnings
• I have added tests that prove my fix/feature works

Screenshots (if applicable)