feat: migrate core internet search to support Tavily provider

[tavily-integrations]

Summary

• Added Tavily as an optional, highest-priority search provider in EnhancedInternetSearchEngine
• When TAVILY_API_KEY env var is set and tavily-python is installed, Tavily is used as the first search method; existing DuckDuckGo/Google/Brave providers remain as fallbacks
• Added _search_tavily() method using the Tavily Python SDK with search_depth="basic" for fast results

Files changed

• models/internet_search.py — Added os import, optional tavily import, Tavily client init in __init__, _search_tavily() method, and wired it as priority 0 in _perform_search()
• requirements.txt — Added tavily-python>=0.5.0
• config.yaml — Added tavily server block under mcp.servers with TAVILY_API_KEY env var reference

Dependency changes

• Added tavily-python>=0.5.0 to requirements.txt

Environment variable changes

• Added TAVILY_API_KEY (optional; when set, enables Tavily as the primary search provider)

Notes for reviewers

• This is an additive migration: no existing provider code was modified or removed
• Tavily import is wrapped in try/except so the app works without tavily-python installed
• The Tavily client is only initialized when both the SDK is available AND the API key is set

🤖 Generated with Claude Code

Automated Review

• Passed after 1 attempt(s)
• Final review: The Tavily migration for core-internet-search is well-implemented and correct. The additive strategy is properly executed: Tavily is prepended as PRIORITÉ 0 only when TAVILY_API_KEY is set and tavily-python is installed, with all existing providers (DuckDuckGo, Cloudscraper, SearXNG, etc.) remaining unchanged as fallbacks. SDK usage is valid — TavilyClient initialization, search() call parameters, and response field mapping (content→snippet, title, url) are all correct per Tavily's API. Dependencies are updated in requirements.txt and the env var is documented in config.yaml. No regressions introduced.