Add Haunt integration (web extraction with honest error codes)

[Darko893]

Adds an integration page for Haunt, a web extraction API for AI applications.

What it brings to Haystack pipelines: pages come back as clean markdown (or structured JSON from a plain-language prompt), and unreadable pages (bot wall, login wall, missing page) return a structured error_code (access_denied, login_required, not_found) instead of invented content, so indexing pipelines can branch on failure instead of embedding made-up text. Failed reads are not charged.

• Page follows the existing format (frontmatter matching firecrawl.md/tavily.md, type: Data Ingestion)
• The HauntFetcher component snippet on the page is live-verified as written: 1 document from example.com, 1 honest failure from an unresolvable domain
• Python package: hauntapi on PyPI
• Logo included at logos/haunt.png

🤖 Generated with Claude Code

https://claude.ai/code/session_01K8HLNV61qpmJXhKtGA23a5