feat: add Exa AI-powered search tool

[tgonzalezc5]

Summary

• Add ExaSearchBackend as a new search engine option alongside BoCha and DuckDuckGo
• Wire Exa into the AISearch auto-fallback chain (BoCha -> Exa -> DuckDuckGo)
• Support Exa search types (auto, neural, fast, instant, deep variants), content modes (highlights, text, summary), category/domain/text filtering, and date ranges
• Set x-exa-integration: alphora tracking header for API usage attribution
• Add exa-py>=2.0.0 to pyproject.toml dependencies

Usage

from alphora.tools.builtin import WebSearcher

# Use Exa as the search engine
searcher = WebSearcher(exa_api_key="your-exa-api-key", engine="exa")
result = searcher.search("latest AI research", count=10)
print(result)

# Or use auto mode (BoCha -> Exa -> DuckDuckGo fallback)
searcher = WebSearcher(exa_api_key="your-exa-api-key", engine="auto")
result = searcher.search("quantum computing breakthroughs")

Set EXA_API_KEY environment variable to use Exa without passing the key explicitly.

Files Changed

• alphora/tools/builtin/web_search.py: Add ExaSearchBackend class, update SearchEngine enum, wire into AISearch and WebSearcher
• pyproject.toml: Add exa-py>=2.0.0 dependency

Test Plan

• Verify ExaSearchBackend.search_web() returns valid SearchResponse with highlights
• Verify ExaSearchBackend.search_news() uses category="news" correctly
• Verify auto-fallback: BoCha -> Exa -> DuckDuckGo when backends fail
• Verify engine="exa" routes to _search_exa() in AISearch
• Verify missing EXA_API_KEY returns descriptive error message
• Verify x-exa-integration: alphora header is set on client