feat: support llmConfig BYOK on jobs and retrieval

[suguanYang]

Summary

• Add LlmConfig / LlmProviderConfig / LlmModelsConfig and optional llmConfig on job create, parse, and retrieval.
• Flat root {apiKey, model, baseUrl} for multimodal; models for same-endpoint different model ids; text / vision for different endpoints.

Depends on: Ontos-AI/knowhere#217

Usage

import { Knowhere } from "@ontos-ai/knowhere-sdk";

const client = new Knowhere({ apiKey: "..." });

// One multimodal model
await client.jobs.create({
  sourceType: "url",
  sourceUrl: "https://example.com/doc.pdf",
  llmConfig: {
    apiKey: "sk-...",
    model: "gpt-4o",
    baseUrl: "https://api.openai.com/v1",
  },
});

// Same endpoint, different models
await client.parse({
  url: "https://example.com/doc.pdf",
  llmConfig: {
    apiKey: "sk-...",
    baseUrl: "https://api.openai.com/v1",
    models: { text: "gpt-4o-mini", vision: "gpt-4o" },
  },
});

// Two different endpoints
await client.retrieval.query({
  query: "What is the refund policy?",
  useAgentic: true,
  llmConfig: {
    text: {
      apiKey: "sk-...",
      model: "gpt-4o-mini",
      baseUrl: "https://api.openai.com/v1",
    },
    vision: {
      apiKey: "sk-ali-...",
      model: "qwen-vl-max",
      baseUrl: "https://dashscope.aliyuncs.com/compatible-mode/v1",
    },
  },
});

Wire JSON uses snake_case (llm_config, api_key, base_url) via the existing HttpClient interceptor.

Test plan

• pnpm test