feat(fetchers): RSSFeedFetcher — structured feed parsing

[chaliy]

What

Add an RSSFeedFetcher that detects RSS/Atom feeds (via content-type or XML structure) and returns structured feed entries.

Why

Agents monitoring blogs, release feeds, changelogs, and news sources encounter RSS/Atom feeds. The current DefaultFetcher returns raw XML, which is hard for LLMs to parse effectively. A dedicated fetcher provides clean, structured entries.

Requirements

• Detect via content-type: application/rss+xml, application/atom+xml, text/xml (with RSS/Atom root element)
• Parse both RSS 2.0 and Atom 1.0 formats
• Return: feed title, description, link, last updated
• For each entry: title, link, published date, summary/description, author, categories
• Format field: "rss_feed"
• Limit entries to most recent N (configurable, default ~20)
• Preserve HTML content in entry descriptions via existing html_to_markdown conversion

Design Notes

• XML parsing needed — consider quick-xml crate (same as potential ArXiv dependency)
• Content-type detection should happen early in the fetch pipeline
• Could also detect RSS feed links in HTML pages (<link rel="alternate" type="application/rss+xml">) and offer to follow them
• This fetcher detects by content-type rather than URL pattern — slightly different matching strategy

Tier

3 — Differentiated capability