Add camel ask command and shared LlmClient with Anthropic/Vertex AI support

[gnodet]

Summary

• Extract shared LlmClient from Explain supporting Ollama, OpenAI, and Anthropic (including Vertex AI) APIs with auto-detection from env vars
• Add camel ask command for AI-powered runtime inspection with tool calling (13 read-only tools: routes, health, endpoints, traces, etc.)
• Support interactive chat mode (camel ask without args) with JLine LineReader maintaining conversation context across turns
• Refactor Explain to delegate LLM interaction to LlmClient

LlmClient

Shared LLM HTTP client with three API types (ollama, openai, anthropic) and auto-detection:

1. ANTHROPIC_API_KEY set → Anthropic direct API
2. CLOUD_ML_REGION + ANTHROPIC_VERTEX_PROJECT_ID set → Vertex AI
3. OPENAI_API_KEY set → OpenAI
4. Ollama via camel infra or localhost:11434

Supports both simple generation (for explain) and tool-calling agent loop (for ask).

Ask command

Agent loop: sends user question to LLM with tool definitions → LLM calls tools → results fed back → repeat until text answer.

Tools (read-only): get_context, get_routes, get_health, get_endpoints, get_inflight, get_blocked, get_consumers, get_properties, get_route_source, get_route_dump, get_route_structure, get_top_processors, trace_control

Two modes:

• One-shot: camel ask "what routes are running?"
• Interactive chat: camel ask (enters ask> prompt with conversation context)

Test plan

• camel explain hello.yaml still works with Ollama
• camel explain hello.yaml --api-type=anthropic works with Vertex AI
• camel ask "what routes are running?" --show-tools inspects a running app
• camel ask enters interactive chat mode
• CI passes

Depends on apache#23346

Claude Code on behalf of Guillaume Nodet