Fix RAG chat temperature handling

[ar-amk]

Part of the open Algora bounty for [ISAAC-497] Implement an enhanced RAG Pipeline for Scientific/Research Workflows.

/claim #45

Bounty reference: https://algora.io/isaac/bounties/clq18zr98000ejs0gt0nv7gwu

Summary

• Pass the resolved RAG chat temperature into OpenAIStream instead of silently forcing every streamed answer to 0.
• Add a small server utility for default-temperature fallback behaviour.
• Safely fall back on the default for missing, NaN, infinite, or out-of-range temperature values.
• Add focused unit coverage for explicit, omitted, and invalid temperature values.

Why this helps the scientific RAG workflow

Scientific and research workflows often need controlled answer variance: deterministic extraction for evidence-heavy questions, and slightly higher temperature for synthesis or exploration. The UI already collects a temperature setting, but the RAG stream ignored it. This PR makes the existing control actually affect RAG answers while preserving a safe default.

Verification

From ui/:

npx vitest run __tests__/rag-chat-options.test.ts --reporter verbose
npx prettier --check pages/api/rag-chat.ts utils/server/rag-chat-options.ts __tests__/rag-chat-options.test.ts
npx tsc --noEmit --pretty false
npm run lint -- --file pages/api/rag-chat.ts --file utils/server/rag-chat-options.ts --file __tests__/rag-chat-options.test.ts
git diff --check

Results:

• Targeted Vitest passed: 3 tests
• Prettier passed
• TypeScript passed
• ESLint passed
• Diff whitespace check passed