feat(core): add Mistral AI support (salvage of #796)

[Bartok9]

Summary

Salvage of #796 by @sachinkum0009 — Mistral AI vendor wiring for chat + embeddings:

• [mistralai] section in config / defaults
• MistralAIConfig + load path
• ChatMistralAI / MistralAIEmbeddings branches
• langchain-mistralai dependency
• test config fixtures include mistral

Salvage fix (extra)

Original get_llm_model_direct branch constructed:
ChatMistralAI(model=model, ...) while the parameter is model_name — NameError at runtime. This salvage uses model=model_name (matches openai/ollama/google).

Credit

Salvage of #796 by @sachinkum0009 — rebased to current main with the NameError fix above.

Verification

• Pre-ship checklist 11/11
• Focused re-read of model_initialization direct + embeddings paths

Notes

AI-assisted; human-reviewed. Prefer this over leaving #796 with the runtime NameError.